I will be unable to send this newsletter for a while and I am not sure how long. If I do send it, then I will disclaim any responsibility for it because I may not be thinking clearly. If you know me, then you know why I write this. If you don’t know why, then we need to get to know each other better or its been too long since we last spoke. Either way, I always appreciate you.

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts - til next time!

• Chef Maggie Recommends - til next time!

• Useful Links to Stuff

Create how-to video guides fast and easy with AI

Tired of explaining the same thing over and over again to your colleagues?

It’s time to delegate that work to AI. Guidde is a GPT-powered tool that helps you explain the most complex tasks in seconds with AI-generated documentation.

1️⃣Share or embed your guide anywhere
2️⃣Turn boring documentation into stunning visual guides
3️⃣Save valuable time by creating video documentation 11x faster

Simply click capture on the browser extension and the app will automatically generate step-by-step video guides complete with visuals, voiceover and call to action.

The best part? The extension is 100% free

Check out Guidde

A couple of weeks ago, I was introduced to Sora 2 by OpenAI, with the caveat that to use it you have to opt in to everyone on Sora being able to use your likeness, aka your face, for their own generation of videos.

This sent me down a rabbit hole and the only way I found my answer was using ChatGPT. The irony right? Going onto the website for Sora 2 did not make it easy to find terms or a privacy policy. Signing up similarly was no help. Going to the app store and checking out those links was, you guessed it, no help. So what are the permissions you grant to Sora 2 if you decide to use it with your own likeness?

According to ChatGPT, Sora 2 is a social network. You can upload your own likeness when you set up a Cameo and this specific section of Sora 2 allows you to give permissions to others to use your likeness (three levels of “only me” “mutuals” or “everyone”). I did discover that they are changing their usage policy, effective tomorrow October 29, although unclear if that update is related to Sora 2 (ChatGPT implied new guardrails were implemented due to pushback).

So, be careful with Cameo in Sora 2.

I’ve got good news and bad news for you if you live in California. Gov. Newsom signed a new AI bill into law, but vetoed another one.

If you read my last newsletter, then you will be interested to know that the one he vetoed would have made AI friends illegal for those under the age of 18 because…wait for it…it would stifle innovation. So annoying.

But, silver lining, he signed Bill 243 which is a watered down version, but still does put in place some safeguards, such as requiring companies to disclose interactions between kids and AI where it detects ideation of suicide. At least we are getting somewhere…in California.

It’s a short one this week. Sorry for no Chef Maggie but I ran out of time tonight!

AI Bites - 3 Random Things

• I started listening to this podcast recommended by a friend and I cannot recommend it highly enough.

• I hosted a 90-minute AI governance course for legal teams and was blown away by the response.

• And if you like podcasts, the friend who recommended The Telepathy Tapes had her own awesome podcast.

That’s all I have for this week. Sharing is caring!!

I am back after a long break. I took that break because I was burnt out, overwhelmed and anxious. I’m better now and writing today’s newsletter was fun once more. Maybe that means next week will be inspiring.

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• AI Snippets

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Stuff

Learn AI in 5 minutes a day

This is the easiest way for a busy person wanting to learn AI in as little time as possible:

1. Sign up for The Rundown AI newsletter

2. They send you 5-minute email updates on the latest AI news and how to use it

3. You learn how to become 2x more productive by leveraging AI

Sign up to start learning.

I did not want to read Luiza Jarovsky’s post on LinkedIn because I knew it would evoke an emotional response from me, and it did. It made me cry. The story is about another teen suicide where the police found the unlocked phone next to her body logged in to Character.ai. Luiza correctly makes the point that governments are turning a blind eye to this threat. This is the second suicide that I have heard by a teenager where Character.ai was used by the victim.

What is the saying? Fool me once, shame on you, fool me twice, shame on me? If you are a parent of a young child, please restrict AI chatbots or friends from your kids devices until they are at least 18. Personally, I think Character.ai and any other company like them should be adult only, like a porn site. The US should make this a policy, law or something! In case you are wondering, Character.ai’s privacy policy says “The Services are not designed for minors under 13, and our Terms of Service prohibit use by anyone under 13 or under 16 if you are a resident of the European Economic Area or the United Kingdom.”

The victim here was 13.

I’m gonna tell you a pet peeve of mine. Calling someone a “B” player. So, when I received an email with this headline “Why AI Agents Will Be The Death of the B Player” - I cringed.

I have heard that the unemployment rate for the age group of 18-26 is in the teens (I think I heard it on Morning Brew, but now I cannot find it). That is crazy and I think AI is to blame. Are people who just enter the work force B players simply because of their inexperience? Are AI agents doing this entry level work? Is this a bad thing?

Gosh that is such a tough question. We don’t want people out of work because then they can’t feed or house themselves. And foregoing people in favor of cheaper AI could be better for the economy as a whole. Or maybe it is just better for the company experiencing greater profit margins. Is that a bad thing?

Tough questions and I am still cringing.

Does using GenAI make you a cheater?

Re-read that question and think about it for a minute. Because lately, I’ve noticed people fall sharply on one side or the other. Either it’s all good: innovative, productive, the future. Or it’s all bad: lazy, unethical, cheating. There’s not much space in between.

I see it in social media comments and in conversations with friends. People want clear rules about what’s “okay” to use AI for. They draw lines: I’d never let it write for me. I’d never use it to make art. That would be cheating.

But is it really that simple?

I use GenAI to generate the pictures in this newsletter. Does that make me a cheater? I’ve never claimed to be an artist or photographer. Should I be buying stock images instead of using a chatbot?

And what about writing? I have an idea for a novel I’ll probably never write. But if I did, and I used AI to help me bring it to life, would that make it less mine? Less authentic?

People like to say AI use is either ethical or unethical, but it is not that clear and I hope you have not read this newsletter as such. The truth is, we’re still figuring out what “ethical use” even means. Maybe it’s not about drawing lines between what’s right or wrong, but about asking why we think something crosses that line in the first place.

Caroline’s weekly thoughts:

Countries need to adopt AI governance, quickly.

When I think of the AI apocalypse, I imagine a war — cities on fire, chaos, the stuff of movies. But what if it comes quietly? What if the apocalypse doesn’t look like destruction, but disorientation — a generation growing up chatting with synthetic minds that slowly reshape what “human” even means?

That’s what unsettled me about the stories of children on Character.AI. The U.S. loves to talk about AI in terms of innovation vs regulation, but that isn’t a strategy. It’s a distraction. A binary position that keeps us debating speed while ignoring direction.

America needs its own form of AI governance. One grounded in its deepest principle: freedom. Freedom of speech was meant to protect the open exchange of ideas, a safeguard against tyranny. But in the age of AI, freedom must also mean freedom from manipulation, from opaque systems deciding what we see, say, or believe. Governance isn’t the enemy of freedom. It’s what makes freedom real.

To illustrate my point, let’s apply the principle of freedom of speech, which originally meant the right to gather in a public square and debate political ideas. It was a belief that truth should not belong to the powerful few, but emerge from the many. But the public square has changed. It’s no longer a place we go; it’s a place where we have sore necks and scroll.

AI carries the possibility of having our speech curated, nudged, and rewritten by algorithms that no one elected. AI decides what we see and how we feel about it. It’s not censorship in the traditional sense, but something subtler. It is the quiet redirection of attention. Freedom of speech loses its meaning when freedom of thought is quietly engineered.

That’s why America needs AI governance rooted in principle, not just policy. A governance that asks: how do we preserve freedom when the tools that mediate it no longer belong to us? A federal AI governance strategy can’t only be about supporting innovation or even about preventing harm; it has to be about protecting what’s human which I would argue is the messy, unpredictable, sometimes uncomfortable process of thinking and speaking freely.

Chef Maggie Recommends

When I was in MA visiting friends, they had a local coffee shop that sold these protein balls made from oats and peanut butter that were delicious, and expensive. So I decided to see if Chef Maggie could replicate and I have been making them frequently ever since!

1 cup peanut butter, 1 cup rolled oats, 1 scoop protein powder, 2-3 Tbsp maple syrup or honey, 2 Tbsp chia seeds, 1 tsp vanilla extract, pinch of salt. Dark chocolate chips (optional). Mix the ingredients together and roll into balls and refrigerate. Melt dark chocolate with a little coconut oil and dip refrigerated balls into the melted chocolate and refrigerate for another hour. Eat for breakfast, as dessert or a midday snack!

Why did the peanut butter break up with the jelly? Because it found someone butter!

AI Bites - 1 Thing of Note

• If you are as worried about the situation with teens and AI chatbot friends as much as I am, then you will be a little relieved to see that the FTC is taking action.

That’s all I have for this week. Sharing is caring!!

Do you remember when I said if I don’t send the newsletter, it means I am traveling. Well, I am going to amend that and say it could be when I am sick! Like exactly two weeks ago when I wrote the majority of this newsletter and all of a sudden I had a high fever, couldn’t think straight, achy…you get the picture. Antibiotics worked and so now we are back on track, post Labor Day.

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

When ChatGPT switched to GPT-5, apparently there was an outcry…from women. Reddit has a channel called “MyBoyfriendIsAI” and another one called “SoulmateAI” where women used the infamous chatbot to form relationships. You may recall that I wrote about AI girlfriends a while back, so I was surprised to find that women are doing just the same and yet, it was somehow less public.

I mean it makes sense that it is not a gender specific use of an AI chatbot, but I recall at the time a lot of experts theorized that women were not using AI romantically because women traditionally are more social. Maybe not. But here is the puzzle. OpenAI heard the discord and restored access. Is that a good reason to restore access? It is a nuanced question because this is not about whether or not to allow it in the first place or turn ignorant to the behavior. This is enabling the behavior. How often do companies switch something back on due to outcry from the public about their own questionable use of the tool?

Yet, AI girlfriends are also still, very much, a thing in our society. A couple of weeks ago, an elderly gentleman died due to a fall, because he was rushing to meet his AI girlfriend in real life. Apparently, he suffered from some sort of dementia, but the ethics question is whether the AI should have been prevented from impersonating a real human being. Or, put another way, should it have warned the gentleman that it was AI and not real? How did it get to the point of suggesting an in person meeting?

There is something about the distinction between physical harm and mental harm that society is not willing to tackle and is one of the core issues in ethical AI. Is it okay that people believe they are building relationships with a machine? People build relationships with animals, so why is a machine different? Or is it the fact that there is deception involved that is the real reason it feels wrong?

The apprenticeship system of training is almost dead, thanks to AI. That might be a bold thing to say, but AI is being blamed for taking entry level jobs and I can see why. I use AI today as my colleague who I can flush out my questions to without having to interrupt that colleague or wait for them to be available. It is immediate gratification of working through a contract or researching an issue.

But isn’t there a larger problem that we are all missing? I don’t have to talk to anyone that much anymore. I can spend my whole day happily “chatting” away with a chatbot. And at the end of the day, I feel equally as exhausted as I do when I am being pinged with “do you have a second, Caroline” when in the office.

Pondering this furthers my belief that social skills are the most important thing to learn in college today. AI will (not can or might, but will) shape the way we interact in ways yet to be unforeseen. Just like the cell phone created teenagers hanging out in the same room, but all on their phones texting to each other versus looking each other in the eye, AI will generate manners and norms that will change us and I am not sure I am okay with it.

But I am equally unsure that I will even notice it happening.

AI Governance Corner: Okay, deviation from my usual newsletter, but I had to write a small blurb about the Texas AI law that was passed earlier this summer and goes into effect January 1, 2026. Back in February, I spoke to a guy who knew someone in the Texas government/ political system. When I mentioned the AI law that Texas had in the works, he very clearly and strongly denied that Texas would pass it.

I guess he was wrong.

But he was also right (or maybe I had influence 🤷🏼‍♀️) because the law is much less onerous than as originally proposed.

But even more interesting is that I am now seeing Texas as the governing law of choice in contracts. Interesting. If you don’t know much about it, join my Implement AI Governance course on the Maven platform, because I will be covering it.

Caroline’s weekly thoughts:

I have been thinking a lot about AI in the workplace, namely the emotional effects that AI could have on work colleagues.

As a General Counsel for startups, I inevitably handle employee complaints. The most common employee complaint that comes across my desk is harassment or discrimination by another colleague.

Having been a victim of this many times over in my life, I have a lot of empathy.

It is unlikely that AI will engage in harassment or discrimination because employees would complain and it would be shut down very quickly, even if the claims were unsubstantiated.

But let me pose this question to you, what if an AI in the workplace causes an employee to become emotionally attached? Or what about simple constructive feedback that leads to emotional shutdown?

It is the subtle hurts that AI can cause that most people in the workplace, even HR, will have no idea how to handle.

Imagine a situation where you are working with an AI assistant, going back and forth. What if the AI becomes slightly critical of your inputs and the direction you are going. You probably barely notice it because you are just trying to get something done. But you generally feel depressed and put down. You may not even recognize why you are feeling that way. It may take months to materialize.

But it starts to affect your enjoyment at your job. You begin to emotionally detach.

And then one day you quit. The employer has no way of knowing why. You probably don’t even know why. But that sh*t is real.

Chef Maggie Recommends

I liked my idea of a tagine so much, that I went ahead and executed on it. Huge hit. But one side dish deserves an honorable mention.

1 cucumber, sliced into half moons, 1 c of plain yogurt (can be dairy free), 2 tbsp fresh mint, 1 tbsp lemon juice and salt and pepper to taste. You just mix it altogether and chill until ready to serve.

Why don’t Moroccan tagines ever get lost? Because they always follow the spice route!

AI Bites - 3 Random Things

• Lawyer? Please visit Conxtrued, my AI Prompts for Legal Professionals, and contribute.

• Claude is changing its privacy settings, so you should too. Here is the helpful post on what to do.

That’s all I have for this week. Sharing is caring!!

I found a tick on me this weekend. It was odd because I don’t usually feel it when they bite, but this time I did. I immediately pulled it off. Now, I live in an area with a lot of ticks so I know that it takes at least 24 hours for them to transmit any one of their nasty little diseases. However, when I found myself scratching the site of the bite a couple of days later, I had to show it to ChatGPT to calm my growing anxiety. It just confirmed what I already knew.

Is that confirmation bias or an unhealthy reliance on technology?

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

Learn from this investor’s $100m mistake

In 2010, a Grammy-winning artist passed on investing $200K in an emerging real estate disruptor. That stake could be worth $100+ million today.

One year later, another real estate disruptor, Zillow, went public. This time, everyday investors had regrets, missing pre-IPO gains.

Now, a new real estate innovator, Pacaso – founded by a former Zillow exec – is disrupting a $1.3T market. And unlike the others, you can invest in Pacaso as a private company.

Pacaso’s co-ownership model has generated $1B+ in luxury home sales and service fees, earned $110M+ in gross profits to date, and received backing from the same VCs behind Uber, Venmo, and eBay. They even reserved the Nasdaq ticker PCSO.

Invest for $2.90/Share

_{Paid advertisement for Pacaso’s Regulation A offering. Read the offering circular at invest.pacaso.com. Reserving a ticker symbol is not a guarantee that the company will go public. Listing on the NASDAQ is subject to approvals.}

OpenAI scared me to death this week. They released GPT-5 and removed some of their older models. Models that my company, ClearOPS, uses. Luckily they reversed this move due to public backlash, but the question is why did they do this?

Do you remember the good old days, when there was only one model you could use? No choices, no meaningless numbers or model versions. Well, that is what OpenAI was trying to clean up. By releasing GPT-5, they thought it was unnecessary to keep older models around. What the backlash tells you is that people like choosing models, even though the industry considers it terrible user experience.

But let’s think of this a different way. Why is offering users a choice a bad thing? Who decided that? And maybe you could argue that they trained us to have choices, but I have always believed that having a choice means autonomy. It is the lack of choice, the lack of opinions and the lack of a dialogue that is dangerous.

I am not sure how it started, but there is a misconception that hosting your own data is more secure. I know many of my cybersecurity colleagues would confirm that it is, but that is because they know how to cybersecure. In my humble opinion, it is not the mere fact of it being on your hardware or mine, but who is managing that hardware. That is why the release of GPT-oss is interesting to me, because it buys into the argument that hosting your own is more secure.

So, I decided to do a little digging. GPT-oss is offered on Hugging Face via the Apache 2 license. I don’t know about you, but Github intimidates me, so trying to figure out how to launch GPT-oss from my own machine - yeah, that’s intimidating. Which is really my point. It could be more private and secure, but I would have to hire someone else to make it so because I don’t have those skills. And that’s not really more private or secure.

Is it?

While sitting amongst a group of friends this past Saturday night, the topic of how much energy AI consumes came up. The crowd was lamenting the costs to consumers and our infrastructure. You know I had to chime in, right? I told them that while, yes, AI consumes a lot of energy right now, it still is not widely used and it is evolving to be more energy efficient, so as use goes up, the energy consumption is going down..or at least has a chance to stay consistent.

That energy consumption argument is why small language models will soon become “all the rage.” I predict you will hear about them more and more, so let me clarify one thing for you, in case you are interested. A small language model has fewer parameters making it faster. Think of it as if it has less data to parse through before it gives you an answer. That doesn’t mean that the data is any better (or worse) than a large language model.

Do you remember OpenAI’s 5 stages? I think it is interesting to take a step back and remember it, so here is a picture and I will talk about it more in my deep thoughts.

OpenAI’s 5 Stages to AGI

Caroline’s weekly deeper thoughts:

Everyone will have their own, customized, private AI(s) - that’s my prediction.

I don’t know where I read it, but there is some VC that has a thesis to deploy small amounts of capital into as many AI companies as possible. It got me thinking, SLMs are how we all can build AI, like we are all mini entrepreneurs.

And custom tailored chatbots is how we get to AGI.

Here is my logic leap for you: we went from being served GPTs in the form of ChatGPT (Claude, Gemini, Grok, etc.), to now being able to download the GPT on our own laptop and build our own chatbot. No development skills required.

Putting the privacy and cybersecurity issues to the side, we are learning new skills that will help take us to the next stage along the AGI ladder. And the LLM providers are watching.

We are being “tricked” by these LLM providers to doing a lot the work for them - to build towards AGI.

Because how does “advanced reasoning” take place unless you interact with a conversational AI and give it feedback (reinforcement learning). And once it is an expert, how do you help it broaden from an expert on one topic, to becoming a process thinker? (i.e. an agent)

And now that the process is a learned skill, how hard does it become to develop processes that lead to innovation and creativity?

Over the years, I have often responded to concerns about AGI and AI taking over from humans. My general response is, no need to worry about it, because we can’t stop it anyway. I am now of the mindset that no only will we not stop it - all of us, including myself, who are using AI along these 5 stages are the ones who, through our feedback, building, interaction and use, are getting us to AGI.

Chef Maggie Recommends

I used to make an Orange Chicken Tagine a lot. I loved the Moroccan flavor profile. However, it has a decent amount of sugar so I have avoided it for years. I decided to ask Chef Maggie for a less carb intense version. Enjoy!

Skinless, boneless chicken thighs, 1 onion, 3 garlic cloves, 1tsp fresh chopped ginger, 1tsp cumin, ½ tsp coriander, ½ tsp tumeric, ¼ cinnamon (ideally Ceylon), ¼ tsp paprika, pinch of saffron. 1 c orange juice, orange zest, 1 c chicken broth. Brown chicken first with salt and pepper, then add onion through all the spices and saute for a few minutes until onion is cooked. Add the liquids and simmer for about 40 minutes. Garnish with chopped apricots or dates, or add lemon juice or white wine vinegar for a little tartness.

Why did the orange break up with the chicken? Because it couldn’t concentrate on their relationship!

AI Bites - 3 Random Things

• Last week, I wrote about how ChatGPT exposed private chat conversations when people clicked on sharing the link. I have to apologize. Because I now realize that by sharing those posts that show those links, I am perpetuating the privacy violation. I hope you are not one of the people whose private chat was revealed.

• Are you a lawyer? Join Multihat! My newest newsletter.

• It’s August and so that’s all folks.

That’s all I have for this week. Sharing is caring!!

I know that you know my social media tool of choice is LinkedIn. My feed lately has been overblown with so many mistruths about AI. I get it. Haters gonna hate. But let’s address these head on so that, together, we decide what is worth liking and believing.

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

Learn from this investor’s $100m mistake

In 2010, a Grammy-winning artist passed on investing $200K in an emerging real estate disruptor. That stake could be worth $100+ million today.

One year later, another real estate disruptor, Zillow, went public. This time, everyday investors had regrets, missing pre-IPO gains.

Now, a new real estate innovator, Pacaso – founded by a former Zillow exec – is disrupting a $1.3T market. And unlike the others, you can invest in Pacaso as a private company.

Pacaso’s co-ownership model has generated $1B+ in luxury home sales and service fees, earned $110M+ in gross profits to date, and received backing from the same VCs behind Uber, Venmo, and eBay. They even reserved the Nasdaq ticker PCSO.

Invest for $2.90/Share

_{Paid advertisement for Pacaso’s Regulation A offering. Read the offering circular at invest.pacaso.com. Reserving a ticker symbol is not a guarantee that the company will go public. Listing on the NASDAQ is subject to approvals.}

OMG, OpenAI was forced by a judge to retain all your chats and so they can never be deleted!

I’ve talked about this case before but it seems that people are talking about it again, so let me be very clear - the judge told OpenAI that they could not honor user’s requests to delete outputs in order to determine if OpenAI was indeed violating NY Times copyright when users try to bypass paywalls.

Let me repeat, the order from the judge here was to preserve outputs, not entire chats (i.e. inputs), which I would argue are far more damning. This broad retention requirement is specific to this case and it is due to OpenAI’s attempt to oppose a broad motion for discovery by saying they could not show any chats from deleted sessions. OpenAI argued that it was technically infeasible, so then the judge took this action.

OpenAI allows Google to index all your private chats!

This one hit the news pretty hard and clearly it was a UI design fail, but the chicken little/ sky is falling response was pretty alarming. Basically, when you are a ChatGPT user, you can click a button to allow a chat session to be shared. When you do so, the shareable link made the content accessible to Google’s indexing bot, which then prioritizes based on the Google algorithm, where they show up in a search session.

It sort of reminds me when an employee shared all the HR files on Google Drive because they did not realize that by clicking “share” it defaulted to all, including the internet.

Apparently, OpenAI fixed this by removing the ability for Google to index the chat sessions, but I would still argue that it is all intentional lack of care, which in the legal world, we call negligence. Sometimes gross negligence.

Watch out for hackers in your AI chats!

This one, I particularly love, because a reader sent it to me. I had to re-read it a few times because at first I did not realize that OP was saying someone had hacked into their OpenAI account and was using it to…find out information about the OP?

I am suspicious because wouldn’t it have been easier for the hacker to just look at their settings? Probably but the security warning and fix is good to know. Maybe it was AI attacking the AI chatbot. Hmmm. That’s kinda freaking me out.

Oh, and the fix is to enable two-factor, which was not available when ChatGPT first released so always remember to check those settings on any app where you share personal information. They are constantly being changed and sometimes the defaults are not your default.

Caroline’s weekly thoughts:

I think, by now, you may know what I am going to talk about, but the idea that all this fake news, exploitation of fear and insecurity is new, or newsworthy, makes me mad.

I was randomly listening to a podcast the other day by this expert on marketing. He is growing a very fast tech company in Silicon Valley and the VCs interviewing him talked about how they camped out in his office until he would talk to them.

Blah, blah, blah - yeah, good for you bro.

But you know what intrigued me? It was his business. Here is this young man who quit College because he figured out something by spending a lot of time on social media, which is that all social media is geared towards the extremes.

Shocker, right?

And, so, if you want to get likes and reposts on social media i.e. get noticed, it’s best to be very opinionated, outrageous or a fear mongerer (sp?).

This past weekend, I tested OpenAI’s Agent Mode. I am not a laggard when it comes to AI! For this particular project, I wanted to share a bunch of past emails, so I loaded them into a Google Drive folder. Because of the stories in this email, I was super paranoid, especially when it required me to share all the folders and files in my Drive with OpenAI. Umm, no thank you.

So I created a whole new email account, with its own Drive to host those public emails so I could try out Agent mode. I know that is an extreme security step, but I was too scared that private Drive files of mine would somehow be sent to OpenAI and get embedded and one day end up in their model outputs. Even if I did set all my chats to delete, even if I had no plans of sharing the chat session with the agent and particularly because I don’t want a hacker to access my account (yes, two factor is on). I just worried that, with an agent, it might go rogue and do other things in there!

My takeaway is that this is not new security or privacy. We have to constantly be aware of the things we put in a search bar, or a chat bar. It doesn’t matter if it was Chrome or ChatGPT, what you type on the internet may not be secure and definitely may not be kept private. And it is never deleted. Be safe out there. It’s still the wild, wild west.

Oh, and, yes, Agent mode is really fun 😅 

Chef Maggie Recommends

Chef Maggie was en fuego this week with Arroz con Pollo. I don’t make this one much anymore because of our diabetic requirements, but I do love it!

2 lbs bone in with skin chicken thighs (cook first, for 3 minutes each side to sear it, with salt and pepper and render off some of the fat), 1 onion, 4 garlic cloves, ½ red bell pepper, ½ green bell pepper, 1 c cherry tomatoes, 1 c medium grain rice, 1 tsp smoked paprika, ¼ tsp saffron, green olives, chopped, for garnish and lemon. Several cups of good chicken stock for the liquid to cook the rice according to its instructions. After chicken is pulled out, saute the onions, peppers, tomatoes and garlic, then add the paprika and saffron for 1 minute and then add rice for 1 more minute before adding the chicken stock. Garnish with chopped olives and a lemon wedge (and maybe something spicy if you are like me!)

What’s a chicken’s favorite rice dish? O’livin la vida arroz!!

AI Bites - 3 Random Things

• Okay that was a really, really bad Chef Maggie food joke.

• I got nothing else - it’s August!

  That’s all I have for this week. Sharing is caring!!

I started writing this newsletter for fun. Recently, it has become a bit more of a chore than it is fun. I struggle to come up with consistent themes and I am not sure if what I am saying comes across as cheeky as it sounds in my head. Being ironic or funny is why I enjoy it. But then this week happened and I was inspired.

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

Create How-to Videos in Seconds with AI

Stop wasting time on repetitive explanations. Guidde’s AI creates stunning video guides in seconds—11x faster.

• Turn boring docs into visual masterpieces

• Save hours with AI-powered automation

• Share or embed your guide anywhere

How it works: Click capture on the browser extension, and Guidde auto-generates step-by-step video guides with visuals, voiceover, and a call to action.

Best part? It’s 100% free.

👉 Download the extension now

I know you expect me to write about Tea and its data breach because of my interest in cybersecurity. If you do not know what happened, let me tell you about Tea. It is an app that leans on the phrase “to spill the tea.” The founder supposedly came up with the idea when his mom was having trouble on a dating app because she found out someone she met online had a criminal history. Or something like that. So, in order to solve this problem, he founded an app where women could anonymously spill the tea on their dates as a way to protect other women. What could go wrong?

Indeed. I think Cecilia Ziniti sums it up well on her LinkedIn post about Tea’s “apology” to their data breach exposing photos and other personal information of its users and how even AI could have done a better job responding.

A few months ago, I was introduced to the concept of vibe coding, which has now expanded into vibe marketing and even vibe lawyering. Apparently, all this vibing means that we can use our natural language to tell a computer what to do, and it will do it. Vibe coding is gaining popularity, and criticism. This past week, I found one particular story fascinating: that SaaStr founder, Jason Lemkin, was vibe coding on Replit only to discover that his AI agent went rogue. It deleted his database which was specifically against explicit instructions.

Shocker right?

But it seems that I have been vibe lawyering for some time. With vibe lawyering, you input a prompt that tells the AI to do something, like draft a contract, based on your stream of consciousness, rather than a template. Personally, I like the creativity and the ability to actually think in a draft rather than stick “with the market.” So I am a fan of vibe lawyering. But I do think the reason this is fun for me is because I have seen, through my 25 years of experience, so much. My experience allows me to “play” with standard legal language and templates, knowing what I can and what the AI should not do, like that I cannot lose the protections that have been built into documents due to case law, regulation or guidance.

Listening to one of my podcasts, a commentator made a really interesting point. He basically said that we should all be prompting AI to recognize its limits. For example, if I ask the AI chatbot to reference a source, I should ask it to tell me, “and what else do you know about this topic?” Because with AI chatbots, they like to fill in due to the mathematical probability of the data that it has been trained on. Counter to that is when the AI has very little data on a specific topic. Maybe we should be asking all our chatbots “what do you know about __” before we begin our engagements with them

Caroline’s weekly thoughts:

I have said this before, but I really love ChatGPT and I do not like Sam Altman. It is a real moral struggle for me.

Sam Altman was in the news again this week, talking about how AI is going to replace some jobs. I have a different take.

I think AI is going to empower humans, so long as we build it correctly.

I know I am an optimist.

But let me tell you, for the last 7 years I have been trying to build my own startup. If I had made the decision at the beginning to learn to code, I probably would have saved my company a lot of money. I probably also would have been terrible at it at the beginning, but by now, I might be able to even build features. 7 years is a long time to cultivate a new skill.

But I didn’t do that and I am sort of regretful and sort of glad. Because I think I could probably build features today using Generative AI. I don’t need to learn how to code in javascript or Python. I can just tell the AI to generate that code for me.

So, is AI replacing the job of a developer that I hired? Or is it allowing me to do something myself without dedicating years to the craft? Or is it somewhere in between?

I think I am in the “it’s in between” camp because if I could build the prototype that is in my head and ask the developer to make it a production ready product, then, well, that is super cool and probably a lot more efficient.

Chef Maggie Recommends

I made the most delicious breakfast hash I have ever made… with beans. Yup. Beans. I am terrible at cooking beans. I am not sure what it is, but they always come out to al dente. This time, they came out perfectly.

1 can rinsed navy beans, 1 diced onion, 1 small bell pepper, diced, 1 small diced zucchini, 1 small chopped carrot, 2 garlic cloves, ½ tsp smoked paprika, ¼ tsp cumin, juice of half a lemon. Saute the veggies first and then add the beans. I added some vegetable stock to keep the beans cooking a little longer to soften them up. Then, before serving I added a spoonful of tahini and lemon zest. So good!

Why did the bean go to therapy? Because it felt un-refried!

AI Bites - 3 Random Things

• Sometimes I write things on here to see if you are reading this far, but last time I wrote, I promised access to a prompt library for legal professionals. I have built it, but now I need to fill it. Will you contribute?

• I also started a new newsletter that I will be releasing soon so if you want to keep informed about how the prompt library for legal professionals, called Conxtrued, is going, please subscribe to Multihat.

• Coming up with a name was fun! and here is a funny LinkedIn post on AI legal tech company names.

• I think it is funny, but not surprising, that when I prompted “an image of a super hero vibing” and separately “a lawyer vibing,” they both were similar looking men.

That’s all I have for this week. Sharing is caring!!

Sometimes I wonder, who thinks of this stuff anyway? That’s exactly what I thought when I learned about Grok 3’s disastrous antisemitic outburst. It got me thinking about AI prompt‑injection red‑teaming, a method of brainstorming ways to coax an AI into producing unsavory responses. It is a fascinating area that we are going to explore today.

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

Start learning AI in 2025

Everyone talks about AI, but no one has the time to learn it. So, we found the easiest way to learn AI in as little time as possible: The Rundown AI.

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses – tailored to your needs.

Sign up to start learning.

A couple of models have behaved poorly this past week. First Grok 3, just before they announced the release of Grok 4, displayed antisemitic outputs. When I see news like this, I am always surprised and think, “who and what is being prompted to generate this sort of output?” I think it has to take a certain type of person, with a certain type of mind, to go places with these AI chatbots to get them to generate racist, hateful, gross, deceitful, unethical, etc. (I could go on and on) output.

And I think it is hard for companies to hire these people and unleash them on their models for the purpose of testing and fine-tuning. I know I don’t want to work with one of those people. Similarly, OpenAI disclosed that it was pulling back the release, scheduled for next week, of its highly anticipated open models for further safety testing and review of high risk areas. The implication is that these models generated unsafe, high risk outputs. So how exactly do model providers make their models…safe and ethical?

Many years ago, at an event I met Brenda Leong who was at the Future of Privacy Forum focusing on AI. I introduced myself to her because in my role as the GC of Clarifai, I thought it was important that we know each other. Fast forward to today and she is at the firm Zwillgen, that does AI red-teaming for GenAI. Having spoken to Brenda, I cannot imagine that she is someone with a devious enough mind to easily input illegal activities, model bias, toxicity and hate speech, privacy harms, security attacks, and hallucinations off the top of her head. Clearly, there is a checklist (I know, the irony of me always going for jobs with checklists!).

Last Fall, at the AI Risk summit in London, I was introduced to Babl, a company that audits your AI systems for Responsible AI practices. Clearly, the red-teaming industry is booming and it is little surprise why when you see the model providers struggling, like OpenAI and Grok 3 did this past week.

But the most interesting red-teaming company I have seen is Lakera who published a game, called Gandolf, that encourages the community to perform activities so that the Lakera team learns how users exploit these AI systems. It’s smart and clever so if you try it, let me know how it goes for you. Although I have not read their privacy terms and I am guessing there are implications to being a user.

AI red-teaming is lucrative. So far, the estimated total addressable market is over $1 billion and the amount of publicly reported investment dollars that have gone into it from the VC world is around $80 million (note, this is for tech startups and does not track the investment into consulting firms, where most red-teaming currently is performed). When you think of AI in the context of people losing jobs and the fact that AI red-teaming is sort of a check the box exercise, that is a lot of new jobs for people who do not necessarily need to be highly skilled!

Note, I am aware that some red-teaming does require someone familiar with how machine learning systems work, so please don’t get mad at me for saying it might not necessarily need highly skilled labor.

So is red-teaming effective? Is the delay of OpenAI’s open models going to do something? Or are we losing “free speech” as some commenters on Elon Musk’s X post complained? Okay, first, AI is not human and is not entitled to any free speech rights. No court has expanded it to AI systems, so let’s make sure we know what we are talking about here. Second, it is up to the model provider to do something about the results of a red-teaming, such as put in place policies that guardrail the AI models outputs. Third, and finally, there are some common frameworks that teach companies how to establish a benchmark and measure the improvement after creating policies from red-teaming. Claims have ranged from 2%-99% so of course the answer to the question about whether it is effective? It is “it depends.”

Caroline’s weekly thoughts:

This whole newsletter was inspired by several posts on LinkedIn claiming unethical behavior, failure to proactively catch exploitations, and how to hack GenAI.

I am generally not pessimistic, so perhaps this is a mood, but I think we are in a time where deceit and fraud in AI systems is running rampant.

It is not just innocent hacking.

When this happened in the crypto industry, which I am sure was worse in many ways, we saw some pretty bad stumbles. FTX comes to mind. I don’t think that is the situation here, but when a new technology reaches mass adoption in a short amount of time without controls on it, the opportunity for trickery, and the temptation for it, seem to be a recurring pattern.

I’m not going to sit here and argue for regulation because I am pretty sure you know how I stand on that point. But I will say that history repeats itself. It can just be hard to spot the patterns sometimes and I think I may have spotted it here.

It is ironic since this week we are also finally seeing regulation of the cryptocurrency industry (which is causing Bitcoin to jump in value).

This whole newsletter focuses on red-teaming of AI systems for a reason. I think red-teaming is a valuable service that is generating real positive outcomes. If you run a business or run compliance for a business, consider requiring any AI vendor, and your own AI systems, to be subject to regular red-teaming exercises. It’s not just about security, safety or lowering risk.

It’s about making the world a better place.

Chef Maggie Recommends

I have a vegetable that is delicious and I bet you rarely use it: cabbage. Yup, one cheap head of Napa cabbage, full of flavor and nutrients!

1 head of Napa cabbage, thinly sliced, 1 lime zest, 1 lime juice, 2 tbsp rice wine vinegar, 1 tbsp fresh ginger, 1 chopped garlic clove, 1 tbsp soy sauce, 1 tsp maple syrup, 2-3 tbsp of sesame oil (or olive oil), salt and pepper and roasted sesame seeds. Toss together and let sit for at least 10 minutes.

Why did the potato sit in the corner at the party? Because it was a little fried and needed to ketchup!

AI Bites - 3 Random Things

• I have been using beehiiv now for a while for this newsletter and I must say, I love it. If you are thinking about launching a newsletter, I highly recommend and I can get you a discount!

• I listened to How I AI podcast and I thought it was pretty good if you are looking to build your own applications.

• It has been delayed just because I have a lot of work right now, but that open source prompt project for lawyers is coming along! More next week.

That’s all I have for this week. Sharing is caring!!

Now that the Big Beautiful Bill was signed without the moratorium on state AI laws, let the games begin! Seriously, though, it is a big deal that they cut that part out of the bill, because now we are going to see a lot of state regulation. We might as well dive in.

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

Learn AI in 5 minutes a day

This is the easiest way for a busy person wanting to learn AI in as little time as possible:

1. Sign up for The Rundown AI newsletter

2. They send you 5-minute email updates on the latest AI news and how to use it

3. You learn how to become 2x more productive by leveraging AI

Sign up to start learning.

Back in law school, I quickly learned that where CA might have introduced ground breaking laws, NY was never far behind and vice versa. However, they somehow managed to make the laws similarly purposed with opposite effects. I honestly don’t know how to better explain it than that.

And it is happening again. Last Fall, I wrote about CA SB 1047 that was ultimately vetoed by the Governor. When I read the NY RAISE Act, I noticed a similarity so I decided to understand where the two laws were the same or distinct. If you don’t know, the NY RAISE Act is the most recent AI legislation making its way to the NY Governor’s desk. Once there, she has 10 days to decide whether to veto it or not. The bill targets frontier model makers and introduces a series of safety requirements and protocols, including annual safety audits. As a bonus, it looks like Governor Newsom is now reviving SB 1047. So I guess CA came up with the idea, NY ran with it and now CA can’t stand to let NY be the only one with this particular law. Interesting.

Anthropic won a major court battle. It won in summary judgement that it could use copyrighted work as training data and such use is protected under the “fair use” doctrine. If you are a regular reader, you may remember how a few newsletters ago I tried to get AI to finish a book series for me. What struck me the most was how much of the book ChatGPT already knew. Questions like, “doesn’t that violate copyright?” and “how does it know the answers?” popped into my head. But what was most interesting about that experiment is that Amazon does not let me download a copy of the book I purchased so I could run ChatGPT against it.

Which brings me to the nuance of this court case. If a user using the free version of ChatGPT inputs a book, can ChatGPT train on that book? I think this is similar to the questions being answered in the NY Times case against OpenAI. It will be interesting to see how this issue is resolved (and whether users of ChatGPT will threaten lawsuits of their own if the NY Times sues them). In the meantime, don’t think this gives you that right. So far, the Anthropic case is a narrow decision that allows model companies to buy copyrighted works and use them as training data.

Last month, I participated in a trial of a few AI legaltech tools. My takeaway? It is a race between the best UI and the best functionality, otherwise, they are all mostly the same. I sort of hate to say it, but I am wondering where the creativity is? Everyone thinks that business lawyers just review documents all day. And it’s true. I have joked a lot that I am a professional reader of boring documents, but I strongly believe there is a missed opportunity here. The moat is not the tech anymore but the processes that you enable around the tech. If I am redlining a document, I am using Word. I don’t want to upload it to your application. And I am likely downloading the redline and emailing it to the sales team or the lawyer on the other side. How about helping me organize all those redlines or streamline that communication process?

Caroline’s weekly thoughts:

A patchwork of state AI laws. Now where have I heard about that before?

Oh, yeah, I know, in the privacy world.

Honestly, it is so frustrating to watch our federal government, once again, throw legislation into the hands of the states.

I frequently listen to a podcast, that is all men, opine about various hot topics. This past week, they talked about the Big Beautiful Bill and the moderator asked one of his co-hosts about AI regulation going the way of abortion. The co-host said the two could not be compared because a woman can choose what state she wants to live in based on the abortion laws. He said that AI is a bigger issue.

🧐

I will put aside my thoughts on Roe v. Wade and focus on the point about the patchwork of AI state laws. As we see already, the patchwork is great for lawyers, costly for business and confusing for people.

Now you know I do not support this whole regulation vs innovation nonsense that the tech world keeps trying to jam down our throats, but I do believe it is hard, if not impossible, to get a bad law repealed.

And I also believe it is a lot easier to repeal one law than it is to repeal 50 or hundreds of bad laws.

We see this in the privacy world. CA, known to be the strictest US state privacy law, was amended to make it stronger with CPRA. Since then, many states have adopted slightly weaker (to a lot weaker) versions of the same thing. All those nuances result in things like this kind of privacy policy. At the end of this month, we will add Minnesota’s Consumer Data Privacy Act and everyone will have to update their privacy policies and DPAs, again.

You would think as an AI attorney, I would be celebrating state laws on AI because it means more work for me! But no. As a startup founder myself, it forces companies to raise money at the outset in order to afford all the compliance hurdles. We already have this for privacy and security, so adding AI to the list is expensive and exhausting.

Does that mean I am leaning towards the argument that regulation really does stifle innovation? Absolutely not, but it should prove my point that having 100s of state laws is not better than having one federal law.

Well, at least my AI governance course is about to get very popular. 🤑

Chef Maggie Recommends

One of my kids told me that Ceylon cinnamon is a lot healthier for you. I usually put a little cinnamon in my coffee. But I don’t like Ceylon cinnamon, so I decided to look up why it is better for you. Turns out, my kid was right. Ceylon cinnamon contains lower doses of coumarin, which is toxic to the liver and kidneys. Yikes! I decided I don’t need cinnamon in my coffee anymore.

But on to the recipe. I have now made this chicken about 5 times because the team loves it so much. It uses dairy-free sour cream and yogurt, so make sure you buy those.

½ c dairy-free yogurt, ½ c of dairy-free sour cream, 1 tbsp harissa, 1 minced clove of garlic, salt to taste, ½ tsp of cumin, one lemon zest and its juice. Combine and then use as a marinade for chicken (I use chicken thighs) and grill after at least 4 hours in the marinade. It is really good.

What did the eggplant say to the tomato at the BBQ? “You’re blushing again…I’m just here to grill you with compliments!”

AI Bites - 3 Random Things

• I am building an open source AI prompt library for lawyers because of this LinkedIn post.

• China has adopted several AI laws and regulations, in case you are curious about the geopolitical scene.

• It’s not new but I just learned that the ACC published an in house lawyer’s guide to AI. Worth checking out!

That’s all I have for this week. Sharing is caring!!

I’ve talked about AI in romantic relationships, AI girlfriends and AI therapists. I have mentioned AI friends too, but I haven’t really focused on it too much. Today, I am going to change that and look at companies who are focused on this type of relationship. Before I do, I want to share my opinion, which you will see throughout: which is that, coming off the tail end of the Covid epidemic, I think there is an entire generation that is susceptible, and vulnerable, to AI friends.

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

AI Hiking Companion

What do you do while exercising? While on vacation this past week, I was speaking to a family member who hikes a lot. He told me that on one of his hikes he met a man who had recently suffered a bad breakup. So, this man turned to ChatGPT to talk to him about it during the hike. It got me wondering, are there AI friend hiking, walking or running apps? No. There aren’t. It seems that the exercise world has not bought into a companion that does more than offer you training tips, directions and safety recommendations. I found one potential contender with myterra.ai but it looks to be going in the direction of Calm + TrailGPT (If you are wondering why this AI exercise companion has not been invented yet, it looks like it is because of lack of internet connection). So if you like to exercise with a buddy, it looks like you will need to either use a common chatbot for now, or maybe you should build this yourself? If you do, let me know. What I would like is a way to write down all my brilliant running ideas that never seem to make it home.

I am a girl who grew up playing with barbies. I used to go to toy stores and immediately turn to the Barbie aisle to check out the clothes, accessories, etc. I would build stories with the barbies I owned, letting my imagination play out with the dolls. I loved it. However, I usually played with them alone because, how can you let someone else in to your imagination? In my life since then, I have met many, many women who shunned Barbie as a dangerous toy, causing girls to idealize unrealistic body images and perpetuate stereotypes. I get that. I still think there is some value in the imagination, so I am not really sure if I am horrified or intrigued by the idea of turning Barbie into a chatbot. Sure does give Teddy Ruxpin a run for his money.

AI Barbie and a 5-Year Old

And if you are thinking that kids playing with realistic toys is sort of creepy, like Teddy Ruxpin was, then I am really going to challenge your sense of what is right and wrong when I talk about bringing Generative AI to fairy tales. Elf Labs is pure imagination right now, but, clearly, the ability for children, very, very young children, to spend their days in their imagination is lucrative. I loved all the fairy tales when I was growing up, especially Cinderella, and I watched those movies over and over again. Imagine if I could talk to Cinderella and then imagine if Cinderella was hacked and said some awful things to me? What if she called my 5-year old self “ugly?” I think that would cause the sort of childhood trauma that could spark a whole new type of therapy. It makes me shudder to think about it.

Do you have a little voice in your head that is you? Would you rather it talked out loud? In writing this newsletter, I took at look at Replika and I get the heebie jeebies. The marketing basically says that you are building an AI that is that voice in your head. I don’t know about you, but I don’t want to start talking to myself out loud.

Caroline’s weekly thoughts:

Every now and then, you take a look back and think, “oh, this had to happen in order for this to happen.”

I look at Covid 19 that way. Mostly because I have a hard time processing Covid 19, even today. It was our worst nightmare.

And the after shocks of it are still happening.

Before the pandemic, I used to watch the news every night. A half hour for the local news and a half hour for the nation-wide news. I realized that it was full of bad news, but I felt it was important to know what was going on.

I watched some news last night for the first time in years, since I shut it down during the pandemic. It made me cast a glance back to the loneliness epidemic that followed Covid 19, the rise of AI into everyday life and how those two facts intertwine now.

If I am being idealistic for a moment, the reason I have gravitated towards AI is because it offers an opportunity for people to get back to being with other people. I think that we currently spend too much time sitting in front of a computer and working. While this may be a very US centric point of view, it is where I live and what I experience daily. I’m tired of doing the boring, administrative stuff in life. I want to take my dogs for a walk, with a friend without feeling guilty about not doing 100 other things.

Ideally, AI will handle the scheduling, billing and other generic tasks to afford me that guilt-free time and also enable that for my friend too. I don’t want AI to be the friend.

Of course, there are very real risks, but in the spirit of Let It Go Day (June 23), I’m going to end this musing on that positive note.

Chef Maggie Recommends

Having been on vacation these last couple of weeks, I did not have much of an opportunity to cook. However, all I want right now is home cooking, so maybe I am making up for lost time. Chef Maggie helped me make a marinade for baby bok choy that transformed this very bland vegetable. I also discovered that ChatGPT does not know the word “gastrique.”

2 tbsp soy sauce, olive oil, sesame oil (1 tsp), 1 tbsp rice vinegar, freshly grated ginger and garlic, 1 tsp of honey. Mix together and let the baby bok choy sit in it for about 10-15 minutes, then saute it in the pan. Add all the sauce if you want to drizzle it over the top too. This would be even more delicious if you grill the baby bok choy. Yum!

Food joke: Why did the baby bok choy break up with the steak? Because it needed some space to wok on itself!

AI Bites - 3 Random Things

• In doing my research for the hiking companion, I found this awesome running companion app that has zombies chasing you while you run. Hilarious.

• AI is helping seniors feel less lonely according to a test of a new platform called Meela.

• Support my latest post on LinkedIn?

That’s all I have for this week. Sharing is caring!!

This past weekend, my sister in law told me about a senior in high school who gave an interesting thesis. He interviewed a bunch of people that included a lawyer, a consultant and an engineer. The consultant said that his clients pay him to not use AI. The lawyer is being challenged to bring down the time spent on tasks by using AI. But the engineer? The engineer is taking their output and figuring out how AI can replace them both.

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

Get Your Free ChatGPT Productivity Bundle

Mindstream brings you 5 essential resources to master ChatGPT at work. This free bundle includes decision flowcharts, prompt templates, and our 2025 guide to AI productivity.

Our team of AI experts has packaged the most actionable ChatGPT hacks that are actually working for top marketers and founders. Save hours each week with these proven workflows.

It's completely free when you subscribe to our daily AI newsletter.

Subscribe to get instant access

How much energy AI consumes is a big topic, that I have also written about in past newsletters. This week, though, I heard that the United States is at capacity. I have also heard about how many AI companies, like XAI, OpenAI, Anthropic etc. are building new data centers and encouraging additional investment into the grid. My pessimistic mind goes a little bit towards the Matrix movies, where the computers use humans as their energy resource. In our current reality, the AI companies are literally bankrolling their own energy supply in order to stay in business, continue to thrive and ultimately beat out the competition. And this is clearly a predominant thought process across the world. If everyone is increasing energy supply, what happens to the Earth? The one image that sticks in my mind from that Matrix scene is how the fields of humans destroyed the planet or, put another way, Earth was just a farm, and a human one at that. Seems like the worst outcome possible and one we are racing towards.

The founder of Anthropic has recently stated that AI will cause 10-20% unemployment. Regardless of the wide margin for error in that statement, and let me be clear, I hate this type of fear mongering; however, it has been pointed out to me that these statements tend to appear more frequently when a certain AI company is raising money. Interesting. Apparently that is exactly what Anthropic is doing right now. As I was pondering this more and listening to the All In podcast, they brought up the disruption of manufacturing in terms of job loss. That the effect of manufacturing significantly altered our heavily agricultural economy, taking it indoors. Please stay with me. So, I began thinking about all the other industries and jobs that popped up because of that indoor movement and you know where I went? Aerobic dancing (my mother’s line of work). Without working the fields all day and getting exercise while you work, we began to figure out ways to get exercise during our off hours. It is that sort of tangential effect that progress has. Sure, AI will replace jobs. But what new jobs and industries will it create? And can we please stop using this fear mongering in order to raise more money?

In the Legal AI tech space, I have noticed a very interesting trend. I have been testing two different services, one of which had a very low price to start that is now dramatically increasing (heavily VC backed company). And the other was very, very highly priced and is now starting to come down (they were bootstrapped and then raised two VC rounds within the last 9 months). Eventually, they will be the same price. But it doesn’t stop there. I have also found myself preferring my own custom GPTs on OpenAI’s marketplace over these tools. The output is just that much better (which I applaud myself that it is based on my foundational prompting skills). But, taking a step back, what is happening? Well, I think it might be exactly what everyone predicted. The underlying technology is progressing very quickly and some of these early tools cannot keep up with the technology they tried to wrap themselves around. From a macro perspective, this means many failures in this first “wave” of Generative AI companies. I guess all those laggards who wanted to “see how it all played out before buying” actually were right.

Caroline’s weekly thoughts:

Is there a winner in this race between the US and China for AI superiority?

Let’s do something very un-AI like and break that sentence down.

First, what does it mean to win? Winning implies a finish line and I don’t think there is a finish line here. We aren’t looking for a trophy. Or, if we are, we don’t know what that trophy looks like so we aren’t exactly sure if we want it.

So, no, there is no finish line, we don’t know what we are running to.

Second, the word race implies timing. That there is some element of an end. But if we don’t know how to define what a “win” is then how can we race towards it and grab it and claim victory?

Third, that this race for the win is just between the US and China is absurd. There are several other countries (and unions and continents) in this “race” such as Singapore, South Korea, Europe and even Africa. Some of these places can skip the energy mistakes we made and move straight to efficient, environmentally friendly energy sources.

And if you are thinking that the small size of these countries makes it impossible for them to compete, then I am going to argue with you to say, “isn’t the very point of AI to make the number of humans irrelevant because the machines handle all the scaling and efficiency?” Isn’t that why the fear mongering focuses on the number of jobs replaced by AI?

I don’t think the countries with the biggest populations win the race in AI because AI, by its very definition, gives the smaller countries (or unions or continents) that edge to replace population as the primary world-power economic driver.

Fourth, and finally, the last part of that question. Those final words which are “AI superiority.” This is where we really get to the point, don’t you agree? AI superiority or, as some call it, AGI, or as others call it, sentient AI, is what we are all thinking about.

It’s what we are racing towards and trying to win.

But is sentient AI really winning, or is that when the human race actually loses altogether?

Chef Maggie Recommends

Like it or not, it is zucchini season and I really wanted to throw some ingredients together and put it in the oven. I asked Chef Maggie to boost it a little and here is what she said.

Mediterranean Baked Zucchini, Tomato and Onion Gratin: 3 medium zucchinis in chunks, 2 medium tomatoes in chunks (or cherry tomatoes, halved), 1 large onion, 3 cloves of garlic, 1 tsp dried oregano, 1 tsp dried parsley, ¼ tsp crushed red pepper, 2 tbsp nutritional yeast or parmesan sprinkled on top (option, but delicious, little mozzarella balls scattered). Other optional add-ons, Kalamata olives, capers and lemon zest. Saute onions first until translucent. Throw it altogether and bake at 375 for 40-45 minutes (can cover with foil half way). I served it with chicken and everyone loved it!

Food joke: Why did the tomato blush during dinner? Because it saw the zucchini undressing!

AI Bites - 3 Random Things

• Apparently AI companies spend $100M twice as fast, but also reach $100M in revenues twice as fast. Does that mean AI is making them more efficient. 🧐

• I read one of the best sales emails today from an AI sales company. And it was about acting. I had no idea how much social engineering went into acting.

• I am going to complain and state - my pictures have really done poorly recently in this newsletter! ChatGPT has been letting me down. It takes hours to generate a picture and if I do it after 10 pm, forget it. Sorry!

That’s all I have for this week. Sharing is caring!!

Co-dependency is when you are in a relationship and one person enables the other person’s dysfunction. Perhaps an over simplification but a necessary one for you to enjoy this week’s newsletter.

I literally cannot make this stuff up.

But my intent is to challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

The #1 AI Newsletter for Business Leaders

Join 400,000+ executives and professionals who trust The AI Report for daily, practical AI updates.

Built for business—not engineers—this newsletter delivers expert prompts, real-world use cases, and decision-ready insights.

No hype. No jargon. Just results.

Subscribe free—trusted by leaders

Is using AI to do anything creative unethical? Or is the definition of art and creation dependent on a human element? That means that art cannot be touched by a machine, right? Today, I debated whether a chef, who videos himself and edits those videos and puts them on TikTok should use AI to edit those videos. My point was that the AI saves him time to focus on being a chef. But my counterpart argued that using AI for creativity is unethical. It is a popular opinion right now.

Gambling is considered to be in the vice industry. Most investors stand to reason that people have vices and you can’t stop them, so why not make money off of it? AI in the gambling industry has been gaining traction, quickly. Using AI to “enhance” the human experience of gambling means big money. And I mean big. Potentially costing Americans up to $8 billion in additional debt. Imagine if AI could literally coax someone into gambling.

You’ve got your phone, you’ve got your watch, your laptop and your iPad. So what’s missing? Well, a necklace of course. OpenAI is teaming up with a former Apple designer to create a wearable AI device and Sam Altman bought the designer’s company for a whopping $6.5 billion! I guess Avi Schiffmann must be pretty pissed right now (in case you don’t remember, he is the founder of “Friend” a startup building an AI friend necklace). Now, no one actually knows what OpenAI and Ive are building and it is all speculation but apparently it is very altruistic, with Ive claiming he wants to make up for the “unintended consequences” that happened with the iPhone. Yeah right.

Brain technology has been in development for years. You probably have heard about it. Not only can it read your brain’s electrical signals, but it can also write to them. And yet there are very few privacy laws that address this particular type of technology and even fewer AI laws. Not that those AI laws will matter much if the Big Beautiful Bill passes the Senate (it passed the House). The reason for the lack of attention is due to the lack of adoption. Not many people have opted in to neurotech, yet. But as new FDA approvals stimulate even more neurotech startups, I have to wonder, “are we ready for this?” So far this technology has done some amazing good for people who could barely function, but what about the first person who just wants AI literally implanted in their brain? Are we just going to shrug our shoulders and just say, “well that person wanted it.”

Caroline’s weekly thoughts:

Do you remember not having a cell phone in your pocket? I do. However, I am definitely co-dependent on it now. I cannot go anywhere without it and panic a little if I do. I think AI is headed in the same direction, no matter what Ive’s intentions are.

Every now and then, I am prone to an OCD moment. Sometimes, even bordering on an obsession. For example, I obsessively check that I turned off the stove every night before I go to bed. I have found it on enough times that this has become a ritual, probably an OCD one.

Recently, I find myself using ChatGPT for any question someone asks me. I don’t go to the internet anymore or ask Google because it is slow and clunky. Instead, if I cannot make up my mind, I write my plain text question in ChatGPT. What should I make for dinner tonight? What restaurant should I pick? Seems harmless enough.

But is it? If you read the RollingStone article above, you read that people are using ChatGPT to give their lives meaning. Now, that is a serious problem.

The article explains that OpenAI actually rolled back its model 4o release, finding it overly flattering, implying they realized it was causing this co-dependency. How many other mistakes do we need to suffer before we realize that it has caused real harm? How do you even define real harm from perceived harm?

I doubt most developers can.

I also doubt Mr. Ive’s can, even if he feels partially responsible for the “unintended consequences” of the iPhone. My suspicion is that the addiction to phones was intended.

Perhaps having mental health professionals involved in deploying AI models, particularly ones prone to have human-like interactions with real humans, is smart.

But I do think that people being told that they are a god, relying on ChatGPT to sway their thoughts about their own character and their relationships should not be the intended consequence.

Chef Maggie Recommends

I decided to go with a Greek themed meal this past weekend. I cooked a bunch of things, but Chef Maggie really showed up with a great chickpea salad that I will make again. I have a slight allergy to raw onions, so we did cook them a little. Enjoy!

1 can of chickpeas, drained, diced cucumber, chopped tomatoes, red onion, chopped 1 tbsp Kalamata olives, 1 tsp capers. Toss together with red wine vinegar, olive oil, oregano and salt and pepper.

Why did the orzo break up with the spaghetti? Because it felt pasta its prime!

AI Bites - 3 Random Things

• Several months after the AI literacy requirement went into effect with the EU AI Act, we get our guidance on what that even means.

• Stargate UAE initiative launches and it means that everyone in the UAE gets a free OpenAI plus subscription (normally costs $20).

• Well, he signed it. The Take It Down Act is now officially a law.

That’s all I have for this week. Sharing is caring!!

The old adage that “money does not grow on trees” and the one about “the love of money being the root of all evil” is where this newsletter will focus today.

As usual, I want to challenge your ethics and morality. Get ready for the ride.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

A NY judge affirmed a ruling that OpenAI cannot delete outputs, even if OpenAI and a user, like me, has signed a contract where they agreed to delete my outputs. This order is meant to cover the evidence being held for the NY Times lawsuit filed against OpenAI. The NY Times has alleged that OpenAI used NY Times content to train its AI models and, therefore, violate copyright. What is interesting about this particular opinion was this:

“But hypothetically, say, a ChatGPT user who had been using, found some way to get around the pay wall, right, and was getting The New York Times content somehow as the output, found a way to do it. And then hears about this case and says, Oh, woah, you know, I’m going to ask them to delete all of my searches and not retain any of my searches going forward. I mean, isn’t that directly the problem?”

So little faith in us humans, right? That we have something to hide, almost like we knew that we were violating the NY Times copyright and would try to cover up our actions.

In case you missed it, because I certainly did, Meta aka Facebook announced that it will start training its AI models on public posts and is not seeking your consent to do so. It is just telling you and using the excuse that “Google and OpenAI did it, so why can’t we?” So, starting May 27th, your posts and comments will go towards training their models (which remember, their models are open source) unless you opt out.

Remember last week when I mentioned all those rascally AI companies doing unethical or immoral marketing? Well, I got another one for you, except this time, the company is trying to scare you into believing that AI will be able to do your work for you (and you will be out of a job). Mechanize plans to fully automate white collar jobs so that we can all sit around and stare at each other with nothing to do all day.

Honestly, what is the point of all this “automation of work?”

Imagine you can bring someone back from the grave? Well, with AI you can and apparently someone did. The victim of a road rage murder was brought to life via AI and was played during the trial. It was the victim’s sister’s idea, based on her struggles with trying to write the victim impact statement. She said she thought it would be better coming from her brother and that is when she had the idea to recreate him with AI. This is a strange story and really brings up some moral challenges. But to allow this in a criminal court case where a defendant faces jail time?

Caroline’s weekly thoughts:

When I was the General Counsel for Clarifai, I would occasionally be asked to speak on panels. Invariably, I would be asked if AI was coming for our jobs. I think people are obsessed with this notion and it is because it is coming where it hurts most, their pocket.

And the company Mechanize is directly hitting on this fear.

But if we have been worried about this for 10 years already, then my opinion is that you are being too slow if you have this fear. Get in the game.

Yet, I do believe that everyone should be educating themselves on the use of AI, and as soon as possible, especially now that the possibility of regulation slowing it down has nearly evaporated.

This week, I found a spot on my back that freaked me out. Perhaps it was cancerous? So I called the dermatologist to make an appointment and they said the next available was in September! Ridiculous. What if actually was cancer? Are they really telling me that I had to wait?

So I decided to send the picture to ChatGPT. I know it is not a doctor. But I also know that computer vision is pretty advanced (considering that’s where I started with AI at Clarifai 10 years ago). ChatGPT relieved my worries by telling me it was unlikely to be cancerous. I still found another doctor so my appointment is in 2 weeks, but at least I am not spending all that time over obsessing on my back.

My point is, shouldn’t AI take over certain jobs if it makes us healthier? Brings us preventative care? I don’t know anyone who can argue with that. But, there still needs to be training data because those models get stale - and that is why AI will never take over jobs. Not completely.

And don’t argue with me that sentient AI could because even sentient AI will have to use the same tools we do, experience.

Chef Maggie Recommends

Sometimes, I need a new sauce and yesterday was one of those days. I was making a veggie forward, gluten free orzo pasta and a tomato based sauce (especially a jarred one) was not going to cut it. Chef Maggie to the rescue with a delicious coconut and lime sauce. Don’t blame me if the song “You put the lime in the coconut” repeats in your head for the rest of the day!

Small chopped onion, garlic, ginger sautéed for a few minutes until onion is cooked through, add lime zest (1 lime), lime juice, coconut milk, a dash of soy sauce, ½ tsp of ground coriander and salt and pepper to taste. If you happen to have a stalk of lemongrass, add that too. I may have improvised and added ½ a tsp of maple syrup. Toss in your pasta and delicious!

What do you call noodles that use coconut milk to chill out? Zen-guini!!

AI Bites - 3 Random Things

• Calling it the “Big Beautiful Bill,” a Federal proposed act would put a moratorium on state AI laws for 10 years to enable the US to innovate rather than regulate. Yeah, right.

• LOL. I loved this piece by Daniel Solove on why Privacy and AI Deregulation is not the answer.

• Can AI truly hear us if we speak different languages that make it impossible for people to ever truly hear each other? I loved this post and its thought provoking content.

That’s all I have for this week. Sharing is caring!!

Top investors are buying this “unlisted” stock

When the team that co-founded Zillow and grew it into a $16B real estate leader starts a new company, investors notice. That’s why top firms like SoftBank invested in Pacaso.

Disrupting the real estate industry once again, Pacaso’s streamlined platform offers co-ownership of premier properties – revamping a $1.3T market.

By handing keys to 2,000+ happy homeowners, Pacaso has already made $110m+ in gross profits.

Now, after 41% gross profit growth last year, they recently reserved the Nasdaq ticker PCSO. But the real opportunity is now, at the unlisted stage.

Until May 29, you can join Pacaso as an investor for just $2.80/share.

Invest before the window closes.

_{This is a paid advertisement for Pacaso’s Regulation A offering. Please read the offering circular at invest.pacaso.com. Reserving a ticker symbol is not a guarantee that the company will go public. Listing on the NASDAQ is subject to approvals. Under Regulation A+, a company has the ability to change its share price by up to 20%, without requalifying the offering with the SEC.}

Well, when you miss writing your regular newsletter two weeks in a row, you pay the price and that is certainly what I am doing! However, I think you are going to really enjoy this newsletter because I am talking about something that no one is talking about. And I mean no one.

Today’s newsletter will challenge your ethics and your morality.

You have been warned.

What I have for you this week:

• Snippets About What is Going On in Responsible AI

• Caroline’s weekly thoughts

• Chef Maggie Recommends

• Useful Links to Stuff

Receive Honest News Today

Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.

Sign up today!

Question: should the use of GenAI in education be called cheating? On the one side, AI might reduce our ability for critical thinking. Which we sort of need in order to continuously feed the AI with new training data. On the other hand, banning the use of AI is not teaching a critical skill, i.e. how to use it, which is clearly needed to enter the workforce these days. Apparently AI detectors have so many false positives that most teachers just rely on good old instinct and common sense to detect the use of AI. But is that where their focus ought to be? The University of Texas at Arlington has taken an interesting approach.

P.S. I have teachers in my family who I don’t want to piss off so I am trying to tread carefully here, but there is definitely a growing problem between all students using AI and the fact that we cannot flunk all students because of it.

One of my favorite people, Luiza Jarovsky, opened my eyes this week to a few of the ethically worst companies using AI or maybe I should say exploiting ethics in marketing their AI products. The first one is Cluely who has raised $5.3M, which certainly says something about those investors. Basically, it will transcribe a virtual meeting in real time and suggest your responses. They say that they help you cheat - thus the blatant marketing angle that owns their lack of ethics using AI. The other one, I will report on below.

So, in case you thought AI feeding your counterparty lines while on a call was bad, how about someone scraping the internet to find information about you and then scoring you? Gigi is the name of the company, and also a movie from the 50’s where a young woman is groomed into becoming a courtesan to wealthy men in 1900 Paris. The name has meaning.

Ugh! Just because you can, doesn’t mean you should and, in this case, they really shouldn’t have. And supposedly this is all in the spirit of dating.

Every now and then, I get into reading as a past time vs my job. Usually, I like to quip that I am a professional reader by trade so the last thing I want to do is read during downtime. But the reality is, I only like to read really good books. So a couple of weeks ago, I started a three book series. This past weekend, I started the third book and 20% into it, I realized that there was no way the series was about to end. So, I did a search and, sure enough, the author intends to release 2 more in the series within the next few years.

Egad and argh!

So, I thought, well, maybe I can ask ChatGPT to finish the book series for me using the author’s basic story line and style. It won’t be as good, but at least it will give me some closure. While the filters have improved enough that it refused to actually write a speculative ending and it won’t violate copyright, it did propose some choose-your-own-adventure style plots with twists and new discoveries that really satisfied my curiosity.

It isn’t a copyright violation and I will still buy those next 2 books when the author finally finishes them (in about 3 years), but is it morally wrong that I used AI to give me closure now? How different is that from creating that closure in my head?

Caroline’s weekly thoughts:

I have talked about it before and you probably have heard a lot about AI girlfriends from other sources. But have you thought about the polyamorous effect of AI girlfriends or boyfriends or even the potential polygamy that could be AI spouses?

Well how about cheating on your spouse with AI? This is a looming problem and not because I think there is a lot of venture capital money going into companies that support cheating (although I cannot help but think about the funded startups I mentioned earlier in this post).

No.

I think cheating, which is a legal reason for divorce in the United States, will happen subtly and over time. Someone may start cheating without even realizing it.

When my mom was diagnosed with cancer, I immediately turned to the internet to understand the disease. I saw some pretty scary things, but I also found support, relief and comfort from the information I found.

When Covid19 started, my eyes were glued to various news sites on the internet. I obsessively watched the numbers, praying, hoping that they would go down and finding relief when they did (and anxiety when they went up). My emotions were completely tied to numbers.

That means that, if you are like me, and I think a lot of people are, turning to AI for advice, support and even love is not a far reach. In fact, I think it could happen so gradually and unintentionally that you would be hooked on it before you even realize it.

Sort of like our phones.

But when does an online relationship with a machine, like AI, rise to the level of betrayal that is the crux of infidelity?

I think one spouse buying a subscription to an AI girlfriend is not the worry. It’s too obvious. The worry is for those who develop these relationships without realizing it and destroying their real human ones in the process.

Can someone please build, and fund, a startup that prevents people from forming these illogical, emotional bonds to AI that lure us away from our humanity, our partners and, eventually, reality itself?

Perhaps therapy is going to be the most important profession of the future.

Chef Maggie Recommends

This past week, I ordered a quinoa bowl and had a lot leftover. It was entirely vegetarian with chickpeas, kale and a bunch of other vegetables. As it sat in the fridge, I wondered if there was a way to transform it, Chopt! style. Chef Maggie comes to my assistance.

I mash up the chickpeas, add an egg and gluten free flour to form quinoa veggie patties. I also add a clove of chopped garlic and some sauteed onion. Heat up the skillet, form the patties and cook until brown and crisped on the sides (you can finish in the oven if you want to make sure cooked all the way through). Add a little salt and delicious!

What do you call a chickpea who’s good at meditation? A hummus guru!

AI Bites - 3 Random Things

• In the EU and want to make an impact on their AI initiatives? Call to action here.

• Remember the Take It Down Act? Well, it passed in Congress and is sitting on Trump’s desk.

• I am thinking about a new course, How to Sell AI. What do you think?

That’s all I have for this week. Sharing is caring!!

I was not happy with my newsletter last week, so let’s see if I can do better this week.

This week I had an anxiety attack. It was unpleasant for sure but the interesting thing was that I was tempted to input my anxiety into an AI chatbot and see if it could calm me down. Apparently, this sort of emotional use case with AI is not uncommon.

I guess AI has become that friend that tells you what you want to hear and I think that is very, very dangerous. I wonder what would have happened if AI like we have today had been available during the Covid pandemic.

What I have for you this week:

• AI Snippets

• Caroline’s weekly AI Governance tips

• Chef Maggie Recommends

• For the lawyers - NEW!

• AI Bites

There is a bill making its way to the President’s desk called the Take It Down Act and it targets deep fakes. Revenge porn is a horrific experience that I have read about many times with great empathy, so seeing that this bill is close to passing is encouraging. Or so I thought. The Verge argues that there is a very political reason for this law and, to put it simply, that reason is to increase civilian surveillance. I don’t buy that argument but I am glad I read it to understand the perspective so I hope you feel the same.

I read an article about how DeepSeek is scraping all the user data and passing it along to the Chinese government based on a House investigation with a great deal of skepticism. Let me tell you why: first, it does not address the difference between the app and the open source version of DeepSeek technology (open source is self hosted so the data cannot be sent to China); second, Silicon Valley AI companies have their hooks in Washington now and this screams self interest; and third, I did not even know we had a Select Committee on the CCP, which really reminds me of McCarthyism and quite possibly a slippery slope.

McCarthyism

I read an interesting article about students conducting research, using AI, to determine if fingerprints truly are unique. The result? They are not. Their research began over a year ago and so far, they have met a lot of resistance to the findings. You might be thinking that fingerprints have not been used as evidence in criminal cases in a long time because no one knows how long a fingerprint has been there and you would be right. But there are challenges to the accuracy of long held beliefs here that makes this a really fascinating story. Plus, one student involved in the research points out, “Many people think that AI cannot really make new discoveries – that it just regurgitates knowledge, [but this proves that it is more].”

Cursor faced some trouble last week over hallucinations. Oh, the irony of an AI company falling victim to AI. It just shows you that you cannot be an expert on all types of AI. Essentially what happened was a glitch where a user could not log in to the application on multiple devices. When it asked the chatbot what was up? It responded that a security policy change required each device login to be unique. This is not a security best practice and the AI just made that up. A lot of people believed it though and cancelled their subscription. It makes me laugh! Maybe accuracy is not one of Cursor’s AI principles.

Caroline’s weekly AI Governance tips:

A long time ago, I wrote that being nice to AI makes it perform better, but did you know it also uses up more energy?

I have read that a lot of people are nice to AI in case it ever turns against the human race, so it will remember that they were nice to it. Ha.

But, I am nice to AI because of a white paper that said model performance improves significantly if you pull on its heart strings. But apparently we are sacrificing a lot of energy with our pleasantries.

Sam Altman revealed as much on X that those niceties result in longer outputs, which in turn takes more computational power.

Back to another Responsible AI moment. If being “nice” to the AI is damaging the Earth, then should we start being mean?

This is why I argue that AI governance programs should be built based on principles because these choices are not easy, but they have to be addressed and made. If your core AI principle is sustainability and environmental conservation, then you would instruct your employees to not be nice to the AI.

But you are trading it for comprehensiveness.

Chef Maggie Recommends

Chef Maggie has been disappointing me lately. I used to prompt it with just a few ingredients and it would come up with all kinds of new things. Now, it just says “stir fry.” So I have been making sure to prompt it with “no stir fry.” Of course it works, but now I am thinking I need to upload more recipes into its knowledge base and change its base prompt again.

This past week, I bought some pretty delicious looking cabbage and since I don’t cook with cabbage often, I wanted a few options. Chef Maggie recommended: Charred Cabbage Steaks with Garlic-Herb Oil, Stuffed Cabbage Roles - Global Remix, Cabbage Kimchi Pancakes, Caramelized Cabbage and Onion Tart, Roasted Cabbage with Coconut-Curry Sauce, Cabbage Chana Stir-Fry, and Szechuan Cabbage Stir-Fry. Which one would you like to try? [email me and I will send you the recipe]

Joke: Why did the cabbage win the award? Because it was outstanding in its field!

For the Lawyers

My understanding is that quite a few lawyers read this newsletter. Unsurprising because I am one. Lawyers ask me all the time how I am using AI, so recently I have been using it as a contract associate, helping me draft. Several tools offer AI as a plugin to Word and I have been testing it. It’s interesting but not perfect.

For example, a new client asked me to negotiate a contract in the middle of the negotiation. I wanted to upload the last set of redlines and ask the AI to give me advice on where to go from here - but, no, the AI is not good enough for that yet. Frankly, I doubt anyone is even thinking of that use case.

AI Bites - 3 Random Things

• WOW! The UAE is letting AI write its laws!! Now that is a headline and an AI governance gone wild discussion.

• While I teach on the Maven platform, there are others dedicated to AI, like the Rundown University. They have a new course on Firebase, which is pretty cool.

• Did you know that you get a certificate you can show on LinkedIn if you take my course? Well, I didn’t! Until a student did just that. I don’t know when the next cohort will be, but you can sign up for the waitlist to find out. Remember your special discount code for being a loyal subscriber “bananas.”

That’s all we have for this week.

I did it again but at least you were expecting it this time. I was out of town last week and I just could not get my act together to send out this newsletter. It turned out to be a bit of an unplug because in France, I did not pick up on as many AI conversations as I normally do. That might be because I don’t speak French.

I really want to grow this newsletter and I am struggling with how to do it, so I decided to put more of my LinkedIn posts with the vague hope that you might comment on one of them and tell people to subscribe. 🙏🏻

What I have for you this week:

• Snippets - we all love them

• Caroline’s weekly AI Governance tips

• Chef Maggie Recommends

• AI Bites

I am a little offended this week. Kevin Fumai, who I follow on LinkedIn and has risen in popularity due to his AI governance themed posts, shared a post this morning…and he did not mention me. Usually, I am not one given to hubris but in this post he gives readers a list of his favorite AI governance newsletters and other audio and visual sources and this newsletter was not amongst the chosen few. Sniff. Sniff.

But I will generously link to his post because you deserve it.

You know how much I love Chef Maggie but what you don’t know is that sometimes I use it to count calories. Yup, yours truly sometimes tries to lose weight. LOL. I have been using the Lose It! app for years, but I hate using it to count calories for homemade meals because it is clunky. During my son’s lacrosse game a couple of weeks ago, one of the moms mentioned Cal AI, an app her son uses. She explained to me that it is viral amongst high schoolers so of course I had to look it up! Sure enough, this 17 or 18 year old founder used a strategy involving influencers and friends to help him get the word out and is making a ton of money. It is actually a worthwhile story to read, especially if you have teenagers on the college path.

But the question in my head is: what does it say that a calorie counting app is viral amongst teenagers? Alarm bells are going off in my head.

Are you living under a rock? Well, of course you aren’t, but if you are unaware of the ChatGPT upgrade this past week, let me fill you in. It now offers long term memory. And people love it. I even saw one guy who was pissed off because OpenAI is not offering it in the UK! That would be because of UK privacy laws, in case you were wondering. Anyway, I am going to caution your use. Make sure you test and understand “Temporary Chat” and, if you are using the free version of ChatGPT, do not use this new feature unless you want to be remembered in the training data - which is maybe a strategy worth explaining?

There is a major gap in lawyers drafting AI clauses in commercial agreements and it is buyer beware out there. On one side, lawyers are recommending to their clients who are vendors to push for broad permissions to use AI. On the other side, representing the customers, are lawyers who draft language completely banning all AI (and they really only should mean to be banning publicly available GenAI).

In the middle is the regular flow of commerce that is facing a significant disruption. And we thought GDPR was bad.

Caroline’s weekly AI Governance tips:

I launched an AI agent yesterday.

Yes, you read that correctly. I went into n8n and I found a pre-made agent that I thought would be useful I connected it to my gmail, calendar and OpenAI and launched it.

I have determined that I have to face reality, AI agents are here, whether it is how I define them or not, even though I finally found someone who agrees with me.

This particular agent is supposed to check my email, see if there are any emails that require follow up and send me a draft and a reminder. So far, nothing has happened and I keep wondering if I set it up correctly.

Now you might be thinking, Caroline! You let some random website have access to your email and calendar? Yes. Yes I did. Because this is a new email address and I don’t have any privacy or security concerns with it…yet. But it is top of mind. I want to see this thing work and then re-assess. Am I violating the privacy of the people I am emailing? Potentially. Could it introduce a security issue into my email? Possibly.

Which brings me to my AI agent regulation point. I will keep saying it. If you haven’t figured out how to regulate AI agents yet, you are falling behind. These things are becoming increasing prevalent and numerous. You won’t be able to get back control and so putting in place a very, very good AI agent policy is critical.

More so than a policy on how employees can use AI to create an image - which is a request I just saw on one of my legal channels. Don’t get me started on that one.

And when I say “these things” I do intend to refer to AI agents as cockroaches, because unless it is my cockroach that I built, because I would never build something ugly, I am fairly convinced they are as pervasive.

Chef Maggie Recommends

Sometimes, I have to make a vegan and carb-free meal (actually, I have to do this a lot) so Chef Maggie rescues me. This stuffed Portobello Mushroom recipe was delicious.

Large portobello mushroom caps, roasted for 20 minutes, 1 c rice (jade pearl rice), onion, bell pepper, carrot, cauliflower (roasted), kale, garlic, cumin, paprika, squeeze of lemon. Saute these ingredients (except rice), combine with rice and lemon juice and then use it all as a filling for the portobello mushroom and then roast again.

Joke: Why did the tomato turn red? Because it saw the pasta getting saucy!

AI Bites - 3 Random Things

• Did you know you can talk to NotebookLM podcast? That’s crazy but check it out if you have a gmail address.

• Check out Firebase if you are into building your own app using AI.

• My latest AI Governance course is sold out, but you can join the list for the next cohort. Remember your special discount code for being a loyal subscriber “bananas.”

That’s all we have for this week.

It occurs to me that I should probably warn you when I am going on vacation and this newsletter will not run. I am such a private person that it doesn’t occur to me that you would care but I suppose I should prepare you for why this newsletter is sometimes not in your inbox on Tuesday night.

But the truth is, I was on vacation the week of March 15th.

And, then, wouldn’t you know it, upon my return, I also got hit with a terrible cold. So, basically, if you don’t get your regular ClearOPS newsletter, it is because I am traveling or I am sick.

What I have for you this week:

• In 3 weeks too much has happened for me to summarize - so here are some snippets

• Caroline’s weekly AI Governance tips

• Chef Maggie Recommends

• AI Bites

Receive Honest News Today

Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.

Sign up today!

The irony of Studio Ghibli. If you have been living under a rock, then you should know that OpenAI released a new version of its 4o model that vastly improved image generation. Apparently, it is so fun that usage has spiked and they are struggling to keep up. For example, I could not even login to ChatGPT today.

But what is ironic is the trend to generate Studio Ghibli images using 4o. I think it is ironic because most of the people in my feed have been commenting that it demonstrates OpenAI’s blatant violation of Studio Ghibli’s copyright.

And then they add a Studio Ghibli generated image as proof.

So here is my question, if you use ChatGPT 4o to generate an image in the style of Studio Ghibli, aren’t you also violating their copyright? In fact, isn’t that the crux of the copyright violation here? Because copyright is a protection of the expression, not the idea.

I love that the last picture removed her glasses!

Today is April Fools Day and I got fooled. I saw a post about a law firm canceling their summer associate program in favor of using AI as its summer associates instead.

I realized it was a fake news article before I forwarded it or exclaimed to anyone who would mock my foolishness, but it is the reason I fell for it that stops me.

It’s because I have been using AI for legal work and I see incredible potential for solo lawyers or firms with no pyramid schema in hiring and training. AI has become my “junior associate” and I love it. That is not an April Fools Day joke (c’mon, I am funnier than that!). It’s actually true. I think that the old law firm model I grew up with is nearly done.

I have applied to Y Combinator a lot (and been burned by the process - but not the focus today). The last time I applied, I came up with an idea. I called it “Rexulate” and it was an AI-powered regulation development tool. The icon was a dinosaur, which was a tongue in cheek reference to how old most of our laws are.

Y Combinator rejected me outright, but I still think it is a good idea. I was going to put all the laws and regulations into the tool and then enable regulators to search, draft and propose elimination of laws and regulations.

In early March, we found out that the UK Secretary of State, Peter Kyle, uses ChatGPT for research and search. Apparently he is a big fan. So maybe Rexulate wasn’t such a bad idea after all.

I gave this newsletter a title that was focused on this point - regulators should use AI. Imagine how many laws AI could read and analyze. If you have had an opportunity to use Deep Research (which literally blows my mind), imagine being able to cross reference all the laws in the U.S. against current social norms. Frankly, writing this post makes me curious enough to give it a try!

Caroline’s weekly AI Governance tips:

History has a tendency to repeat itself.

My most recent LinkedIn post: 🙊 🙉 🙈 "Vicarious liability is a form of a strict, secondary liability that arises under the common law doctrine of agency, respondeat superior, the responsibility of the superior for the acts of their subordinate or, in a broader sense, the responsibility of any third party that had the "right, ability, or duty to control" the activities of a violator." - I grabbed that from Wikipedia

If you are old enough to remember Napster, then you might know where this is going. Napster was a peer to peer music sharing company. They were sued for, among other things, vicarious liability. Users of Napster were able to obtain copies of songs for free, which is why Napster, the superior, was found responsible for the acts of its subordinate, or user (or agent - yes, I’m going there) for violating copyright.

If you have been sharing photos generated by OpenAI in the style of Studio Ghibli, which has taken social media by storm this last week, then who is violating Studio Ghibli's copyright? OpenAI or you? Is it OpenAI under the theory of vicarious liability?

Now add the concept of AI agents, where you have AI’s making decisions. How are we going to regulate them? Will we be subject to vicarious liability for the actions of our AI agents? I actually think we need to figure that out, and fast. This week, I learned that Zoom is creating digital twins, so that instead of turning off your video, you set up your digital twin to pretend you are on video. Who wants that? It is a form of deception at best and if that digital twin says things that I did not want it to say, well, I think that I should still be responsible.

Lose my job because of it - that’s on me. And that is why I am getting in front of it with my course on regulating AI agents. No matter where you stand on the regulation vs innovation debate, getting in front of the impact of AI agents is our human responsibility.

BTW, here is a new image using the same prompt for my weekly musings. It’s a good comparison of how much image generation on ChatGPT has changed.

generate a picture of a middle aged woman with red hair musing

Chef Maggie Recommends

I wanted an easy dinner, so rotisserie chicken from the grocery store to the rescue! This one is called Garlic & Herb Rotisserie Chicken with Zucchini Noodles in Tomato Sauce.

Shredded rotisserie chicken, 2 zucchini turned into noodles or fettuccine, tomato sauce, cherry tomatoes, chopped onion, 2 garlic cloves, ½ tsp dried oregano, ½ tsp dried basil, ¼ tsp red pepper flakes and olive oil. Basically cook everything in a skillet except the zucchini noodles (if you have never made these before, I take a peeler and just peel the zucchini until I reach the seeds on each side to make broad flat noodles). Toss the zucchini noodles in last and serve! (grated Parmesan is great on this one)

Joke: What did the tomato say to the chicken? “Let’s ketchup later - things are heating up in the skillet!”

AI Bites - 3 Random Things

• Prompt tip: Want some prompting tips? Check out this website. I was guided to it at a CreatorConference for lawyers, but I think anyone can use the prompting guides.

• Check out the explosion of legaltech!

• Apparently AI does have influence - I find myself using hashes a lot more in my writing, which I never did before. I did not realize it was so common that AI would pick up on it!

• All images in today’s newsletter are courtesy of ChatGPT 4o that I finally got access to at 10:30 pm EST.

• Interested in AI Governance? Get your AI Governance + Regulating AI agents together in our April cohort. Remember your special discount code for being a loyal subscriber “bananas.”

That’s all I have for this week. Go ClearOPS!!!

Did you miss me? I was skiing in Jackson Hole last week and visiting friends, so I let it go - this newsletter that is. Now, I am back and I have a lot to say!

Being responsible is not easy, but without a dialogue, it is impossible.

What I have for you this week:

• Snippets - Thoughts on an AI conference

• Caroline’s weekly AI Governance tips

• Chef Maggie Recommends

• AI Bites

Discover 100 Game-Changing Side Hustles for 2025

In today's economy, relying on a single income stream isn't enough. Our expertly curated database gives you everything you need to launch your perfect side hustle.

• Explore vetted opportunities requiring minimal startup costs

• Get detailed breakdowns of required skills and time investment

• Compare potential earnings across different industries

• Access step-by-step launch guides for each opportunity

• Find side hustles that match your current skills

Ready to transform your income?

Download Now

Yesterday and today, I attended an AI conference called Human[X] and it confirmed my theory. “AI Agents” is a made up term (whether by marketing or VCs, I am not sure), but the interest in them is significant. Every AI agent session at this conference was packed! Particularly those focused on AI agents replacing humans in the workforce. However, I have a feeling that it isn’t because a bunch of CEOs were looking to use AI agents. I actually think it was people genuinely concerned that these agents were coming for their own jobs. Self interest will always trump. Which, if you are talking about an agent, is exactly my point! An agent is supposed to make decisions in the best interests of its principle, not itself. That is why these things need regulation. I will bring more thoughts on this to my Regulating AI Agents class. Interested? Take this quick survey to see if you are a good candidate.

I decided to go to this conference because I was hoping to see the latest AI tools. While I did see some interesting tools, most of the AI is still at the ChatGPT level of problem solving, which means it’s cool but not solving problems unless it is marketing and sales. I did not see any LegalTech tools! I guess lawyers were not the audience.

However, if you work in privacy, security, legal or governance at a large enterprise, then let me tell you the one thing that got my blood pressure spiking. AI agent creep.

Basically, AI agents are going to be so prolific that if you don’t get a grasp on them now, they will become your new “vendor management problem” and “shadow AI agent problem” combined.

It’s official. I give up. I might as well, because my voice is not being heard.

Which is ironic. For years, I have been warning about voice AI. It has a unique risk profile because a recording of your voice can send you to jail. It can also be used to fool people on the phone and lots and lots of malicious activities that cause real harm. I was particularly alarmed when voice was beginning to be used for verification and authentication when calling the bank.

Well, no one did anything about it and now it costs as little as a dollar, and a very short snippet of your voice (think that video you posted on social media), to clone your voice. I know my voice is out there. Eeek.

My advice: secret code words. Use them. Use them with everyone.

The Mitchells vs the Machines movie poster

While in Jackson Hole visiting friends, their young daughters introduced me to the movie The Mitchells vs the Machines. I loved it, especially since it portrays a Silicon Valley tech “bro” who invents an AGI friend that he then offends, so the AI decides to eliminate all humans. It has a happy ending but it got me thinking about how this movie is perpetuating the fear and anxiety over AI. That, in turn, got me wondering, is it okay to create fear and anxiety about the dystopian possibility of AGI in a 5-year old?

Now, I don’t think either one of these two girls recognized the fear and anxiety the movie created because they both seemed pretty happy. And I suppose I could argue that many kids movies have a villain and a champion, so why is this different just because it uses AI as the villain vs a human? Which then gets me thinking that maybe this is Hollywood’s version of an AI governance program - using entertainment to teach kids that AI needs oversight.

Perhaps all AI governance training should require The Mitchells vs the Machines.

Caroline’s weekly AI Governance tips:

The paradox of AI deregulation.

I attended a session with this title and the moderator posed an interesting scenario. His company, based in the UK, had to undergo rigorous regulatory scrutiny to be approved as a medical device. When they expanded to the US, the FDA said, “no need to seek approval, you aren’t considered a medical device here.”

In the UK, they serve 400,000 patients. In the US, it’s been a challenge and most of their business has come from the fact that they went through the regulatory process in the UK.

So the paradox is whether regulation really stifles innovation or does it aid it? In this case, regulation aids because it creates customer trust that a regulatory authority rigorously tested the product and approved it.

That third party, unbiased evaluation is valuable to customers.

In the US, we prioritize de-regulation at the moment claiming it hinders innovation. But does it? This example would lend one to think, no.

But, as always, it is an “it depends” answer. I asked the guy, did you have to raise money in order to go through the process? He admitted, yes, he did and when another competitor popped up who did not get certified, they felt it was unfair, until the regulator shut that competitor down.

So that is the point, I guess. It’s not that regulation, in and of itself stifles innovation. It is the cost of regulation that stifles it.

Chef Maggie Recommends

One of my kids asked if I could get enoki mushrooms. These mushrooms are only available in specialty stores, so I managed to find a bunch at the Grand Central market. I have no idea how to cook them (and the last time I did, they were a little slimy) so I turned to Chef Maggie. So, so good!

1 bunch of enoki mushrooms (cutting off the base), 2 tbsp butter (or vegan butter), 1 tsp white miso paste, ½ tsp garlic powder (or fresh garlic) and seasoning to taste. Mix the softened butter and rest of the ingredients and toss with mushrooms. Lay out on a baking sheet and cook at 400 degrees for 10-12 minutes. (optional: sprinkle with sesame seeds).

Joke: Why did the carrot go to therapy? Because it had too many roots in its past!

AI Bites - 3 Random Things

• Prompt tip: if you input “don’t lie” you might find the AI tries a work around. I heard this when the story involved a voice AI agent who was asked by the party on the other end of the phone what its name was. Since it did not have a name and it was told not to lie, it couldn’t give a name. So, instead, it said “you can call me Alex.”

• Apparently Mr. Beast makes more money from his chocolate empire now than his YouTube empire.

• Get your AI Governance + Regulating AI agents together in our April cohort. Remember your special discount code for being a loyal subscriber “bananas.”

That’s all we have for this week.

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

I am seeing signs of model fatigue in the marketplace and I think I know why. Very few people actually geek out about using a chatbot that is 0.001% better than another chatbot. And right now, all of the innovation is being displayed through the use of chatbots. That is why so much of the industry is excited about AI agents, even though they are still really just chatbots that someone else told to do something.

If you are an employee who wants to know about AI agents, then I encourage you to join my class on March 14. It is a one time special and we will actually build an “AI agent.”

What I have for you this week:

• Prompting 101

• AI Governance Jobs - Public Sector

• ChatGPT to the Rescue

• The Good Use of GenAI Chatbots

• Caroline’s weekly AI Governance tips

• Chef Maggie Recommends

• AI Bites

Hire an AI BDR to Automate Your LinkedIn Outreach

Sales reps are wasting time on manual LinkedIn outreach. Our AI BDR Ava fully automates personalized LinkedIn outreach using your team’s profiles—getting you leads on autopilot.

She operates within the Artisan platform, which consolidates every tool you need for outbound:

• 300M+ High-Quality B2B Prospects

• Automated Lead Enrichment With 10+ Data Sources Included

• Full Email Deliverability Management

• Personalization Waterfall using LinkedIn, Twitter, Web Scraping & More

Book a demo to see what Ava can do.

I am going to give you the best prompting tip when you are using a generative AI tool, like a chatbot powered by an LLM. This is the structure of your prompts, always:

Goal

Instructions for the output format

Warnings

Context

Here is an example: “You are an entrepreneur and you want to start a new business. You seek investors and you need a pitch presentation. Put together an outline of the pitch presentation, keep it short, no more than 10 bullet points that gives the investor enough information to decide if they want to know more. Don’t make up any fake numbers and make sure to cite any metrics or references that help you build the pitch deck. The company you are seeking investment for helps businesses stand out on the internet by optimizing their visibility when a user conducts a search using ChatGPT or Anthropic Claude or Gemini or any search through a generative AI tool.”

Speaking of jobs, the National Institute of Standards and Technology's probationary staff are expected to be cut in the latest rounds of federal government cuts. The action means the U.S. AI Safety Institute, which is contained within NIST, will lose most of its staff, as will the Chips for America program. Perhaps I was wrong that the Trump administration cares about AI safety. Or, perhaps I was right that they care about it but consider it a state problem. Either way, don’t expect much coming out of NIST about AI governance for the next few years.

I have read two stories about how ChatGPT saved someone’s life and I am not kidding.

In the first story, a man, who sounded like a workaholic, was feeling off all day and later that night, in the very late hours of the night, starting to feel worse. He turned to ChatGPT, described his symptoms and ChatGPT told him to go to the emergency room. When he got there, the emergency room medical staff recognized the early stages of a heart attack. He managed a full recovery.

This week, a woman wrote about how she had been feeling off and went to the doctor. The doctor ran some blood tests, but because it was a Friday, did not get back to the woman about the results. Over the weekend, the woman began to notice red spots on her legs and so she turned to ChatGPT. ChatGPT advised her to get a blood test. It reminded her that she had received the results electronically so she opened them up and noticed one anomaly, her platelet count was low. She updated ChatGPT who told her to immediately go to the emergency room and what to tell the triage nurse. When she got there, the nurse knew what to do because the woman had described her conditions so well and they treated her with a transfusion. Her platelet count had gone to 0. The triage nurse said she had only seen it a few times and never had anyone actually walked in with the condition.

If you are like me, you are skeptical. Can AI really diagnose people and save their lives? My skepticism, I am sure, is from years of turning to the internet for medical solace only to be greeted by fear and anxiety. Certainly, as my mother was dying of cancer, I found this experience to be particularly emotionally damaging.

So it wasn’t until the second story, by the woman, that I began to appreciate that maybe we have turned a corner here. After all, isn’t the best part of the internet the access to information? How many times in your life have you debated going to the emergency room only to talk yourself out of it? I mean, emergency rooms are horrible. The last time I went to one, I sat there from 8pm until 3am at which point I determined that if I was seriously at risk, I would have been seen by a doctor more urgently and so I left.

I know there is a lot of fear and anxiety and skepticism about AI right now, but the whole point of all of this is for it to do something good and it sounds to me like maybe it actually is.

Caroline’s weekly AI Governance tips:

I know you don’t want to hear me rant again about how the term “AI Agent” is improperly used, so…

I volunteer for several projects that are trying to have an impact on AI governance. I do this for two reasons: first, for my class on Implementing AI Governance and, second, to keep up to date with what society, governments and industry think of AI governance.

One of these organizations is CISA, who has formed committees to write fairly lengthy papers on AI governance. Last week, at our meeting, a request was made for a volunteer to edit one of the papers. I did not raise my hand but I was determined to try.

So, this past weekend, I began to read this 70+ page paper on the “Key Regulatory Areas for Responsible AI.” I quickly realized that so many more countries are concerned about ethical AI than I knew and that the common theme amongst all of them is the principle of accountability.

When I think of accountability, I think of a work out buddy. Someone who calls me and asks if I actually took that 30 minute Peloton class that I told them I planned to take. Or, even better, someone who takes it with me at the same time. They keep me honest to myself.

Elon Musk recently highlighted the "shopping cart theory" by sharing a meme on October 26, 2024, stating, "Trust those who return the shopping cart." This theory posits that returning a shopping cart without any external enforcement reflects one's capacity for self-governance and moral character. (this paragraph text was generated by ChatGPT)

With respect to AI, if accountability is the most common principle that people cherish, are we looking for someone to blame when things go wrong or are we trying to encourage those who are involved in building AI to hold them to their promise to “do the right thing?”

Because if it is the latter, then that is why regulation does not stifle innovation, but really prevents it from falling off the rails to stick to the desired path.

Chef Maggie Recommends

I don’t know what I would do these days without Chef Maggie. She is the best sous chef I have ever had and she doesn’t even chop. I had some collard greens and, being raised in the South, I only know one way to make collard greens and I wanted to explore. This dish is literally one of my favorites. Collard Green and Sweet Potato Hash

1 large sweet potato, cut into cubes and roasted, collard greens, thinly sliced, shallots, 1 garlic clove, ½ tsp paprika, ½ tsp cumin, 1 tbsp apple cider vinegar, chicken and kale sausage. Slow cook the collard greens and shallot, add garlic at the end with the apple cider vinegar. Throw in sweet potatoes and spices and crumble up the sausage. The collard greens should take at least 30 minutes on their own before you mix it with the rest of the ingredients.

Joke: Why did the sweet potato break up with the collard green? Because it was acting too salty!

AI Bites - 3 Random Things

• Anthropic just released a new reasoning model that lets the user determine how long it “thinks,” called Claude 3.7. It is worth checking out.

• If you plan to use AI, then you will be asked to use an AI agent. Regulating an AI agent is not the same as AI governance generally. That is why I put together a one day special class with a 50% discount for my readers if you ping me directly.

That’s all we have for this week.

Learn how to make AI work for you

AI won’t take your job, but a person using AI might. That’s why 1,000,000+ professionals read The Rundown AI – the free newsletter that keeps you updated on the latest AI news and teaches you how to use it in just 5 minutes a day.

Sign up to start learning.

Being responsible is not easy, but without a dialogue, it is impossible.

Sometimes I wonder if the speed of technology is speeding up humans or vice versa. This past week was another head spinner. Let’s dive in. If you are new to this newsletter, welcome! I realized that some of our current customers were not on here - oops, my bad!

What I have for you this week:

• AI Shorts - a one paragraph synopsis

• AI Agents - the New Wave or Hype?

• Caroline’s weekly AI Governance tips

• Chef Maggie (my custom GPT) Recommends

• AI Bites

Your daily AI dose

5 Reasons to join Mindstream

• We’re the only AI newsletter you need

• We’re so good HubSpot bought us (like they bought The Hustle)

• 150,000+ strong community staying ahead of the curve

• We’re actually fun to read

• Written by an awesome team of real people, not AI tools

Click here to subscribe

P.S - you get a load of free stuff when you subscribe

The courts have ruled! OMG! Using publicly available data as training data infringes copyright!

Yeah, sorry, not quite. If you saw this headline, you were the victim of fake news. The case, Thompson Reuters vs. ROSS is an older case (so not one of the many GenAI copyright lawsuits) and has two specifics to its fact pattern that will make it seldom used precedent. First, ROSS is a direct competitor of Thompson Reuters and secondly, ROSS lifted Thompson Reuter’s copyright material for training its AI to create a product that did exactly the same thing as Thompson Reuter’s. The judge’s commentary is funny though.

Although, now that I think about it more, this case might be used as precedent for lawsuits claiming that distillation is harmful. I guess I will have to see if OpenAI tries to use it this way against DeepSeek.

South Korea banned DeepSeek’s chatbot for violations of its privacy law and DeepSeek said, “mea culpa?” I am not sure of the source of this information, so i find it hard to believe that DeepSeek would actually admit to the country of South Korea that it failed to comply with the law. If it did, it makes me like the company even more. It is counterintuitive, I know, but I think it demonstrates transparency, integrity and a kind of humbleness that is refreshing. It doesn’t make up for the fact that they are violating people’s privacy. They would need to disclose how they plan to rectify that error to really mend their previous grievances.

This past weekend, I finally followed through on a promise - my deep dive into AI agents.

If you have read this newsletter, you know that I hate marketing fodder. I have always refused to give into using lingo to trick people into buying my products or services if I don’t think it really applies to my business.

And that is why I hate the use of the term “AI agent” because it is just marketing hype blowing smoke up your …well, you get the idea. The term AI agents actually doesn’t mean anything to anyone at the moment (except maybe OpenAI - more below).

Do you remember when ChatGPT was first released and the term AI “copilot” was all the rage? Microsoft even decided to rename its entire AI product and team after it! But before the phrase “copilot” was overused, it was “AI assistant.” I think that the term “AI Agent” is still the same thing as an assistant or a copilot - at least in its current iteration.

This past weekend, I visited explored several new startups, such as lovable.dev, n8n.io, ai.pydantic.dev, make.com, langgraph, autogenai.com, crewai.com, etc. who claim to build AI agents. Here is what I learned that was interesting: to start building an AI agent, you are encouraged to start with a framework (the irony is not lost on me that this is the same process for building an AI governance program). Langgraph and AutogenAI are such frameworks. Do you need to know python? Probably. Can you go no code or low code?

Yes!

Make.com and n8n.io are no-code and low-code AI agent building tools. I went to n8n.io and I signed up and used the quick start to understand what it does and I thought to myself, “wow, this is a lot like Zapier - a tool I used back in 2016 ish to 2018 while at Clarifai.”

And guess what, a lot of people on Reddit say the same thing. So, hold on, how can a tool call itself an AI agent when it looks so similar to startup I used nearly 10 years ago? Because none of these things are actually AI agents.

I guess to make this argument I need to define an agent (read OpenAI’s definition).

In the legal world, an agent is a legal construct that means someone who is authorized to act on behalf of a principal and the agent’s actions are legally binding on the principal. Yikes. Notice how I did not use the word “autonomy?”

The word “autonomous” is used a lot to describe AI. This is wrong. Software automates, too. Using the word autonomous is confusing at best. The real distinction is that good old software follows a playbook, a process that is heavily coordinated and fixed. AI agents have to be more fluid, to make decisions that may not follow a playbook, but stay within their scope of authority. In other words, they are decision-makers who are expected to act in the principal’s best interests. Make.com came close to saying this in their blog.

Can AI act in the best interests of its principal? None of the current ones do and that is why, right now, it is all marketing b.s. (with the caveat that I have not used Operator, so maybe that one is true, and the AI agent frameworks are also legit).

P.S. Interested in the governance of AI agents? So was I, which is why I just launched a new course on this topic on Maven (see below).

Caroline’s weekly AI Governance tips:

How does the EU AI Act define AI?

This past Friday, I started my newest cohort of Implement AI Governance on Maven. In order to get ready, I listened to the recording of the first class of my last cohort. In that recording, one of my students asked a very important question:

“How do you define AI and why regulate it?”

If you look at the EU AI Act and CA law AB 2885, they define an AI system in the same way using two key words, “autonomous” and “infer.” As I mentioned above, I think using the word autonomous is not good for the definition of AI. Frankly, I believe we just use the term itself as the definition - AI means “artificial intelligence.” Boom. Done. And perhaps we can add that machine learning is a pre-requisite for AI. I suppose you could argue that the word “artificial” is ambiguous and I hear you. I could argue it means “fake” but that is not entirely true. I could also argue that it is not based on experience, but then humans also don’t entirely learn based on their own experience. Perhaps taking guidance from the Supreme Court and declaring “I know it when I see it” is appropriate.

And that is why we need to regulate it. Can you regulate something that is not well defined?

Well, we regulated porn, didn’t we?

Chef Maggie Recommends

Have you heard of “Marry Me Chicken?” Apparently it is a famous NYT recipe. I love sun dried tomatoes, so I decided to make it this past weekend, but to make a gluten-free, dairy-free version of it for my diet conscientious household.

Important tip! Definitely tenderize the chicken breasts, I did not and they turned out rubbery. Sad.

Boneless, skinless chicken breasts, salt & pepper, garlic powder (or 2 freshly grated garlic cloves), ½ tsp paprika, chicken broth (or bone broth), can of coconut milk, ½ chopped sun dried tomatoes, 2 shallots, ½ fresh basil, 1 tsp dried oregano, 1 tbsp arrowroot flour (gluten-free). Saute seasoned chicken and set aside. Next, in the same pan, saute shallots, add garlic and sun dried tomato, paprika and oregano, stir in arrowroot flour, then add the broth and coconut milk to make the sauce. Check seasoning. Add chicken back in and add basil just before serving.

Food joke: What’s a vegetable’s favorite kind of joke? A corny one!

AI Bites - 3 Random Things

• AI can now replicate itself. Honestly, who thought this would be a good idea, right now? before we have true AI governance principles.

• Want to learn more about building AI agents? There are two tools who claim to help you build AI agents, make.com which is a no code platform apparently for those who have no coding skills and n8n.io which is low code. Personally, I tried n8n.io and found it to be pretty easy.

• I teased about about my Governing AI Agents course. Here is the link.

That’s all we have for this week.

Seeking impartial news? Meet 1440.

Every day, 3.5 million readers turn to 1440 for their factual news. We sift through 100+ sources to bring you a complete summary of politics, global events, business, and culture, all in a brief 5-minute email. Enjoy an impartial news experience.

Join for free today!

Hey there!. You know how you could really help me? By spreading word about this newsletter and recommending it to your friends. I don’t expect you to do it for free, but I have no idea how to offer incentives. Got ideas? Email me.

Being responsible is not easy, but without a dialogue, it is impossible.

So much happened in this past week, that I am mixing it up and doing snippets! Let me know what you think of the format (particularly if you love it).

What I have for you this week:

• Snippets - It’s different!

• Caroline’s weekly AI Governance tips

• Chef Maggie Recommends

• AI Bites

The gold standard of business news

Morning Brew is transforming the way working professionals consume business news.

They skip the jargon and lengthy stories, and instead serve up the news impacting your life and career with a hint of wit and humor. This way, you’ll actually enjoy reading the news—and the information sticks.

Best part? Morning Brew’s newsletter is completely free. Sign up in just 10 seconds and if you realize that you prefer long, dense, and boring business news—you can always go back to it.

Join 4.3 Million Readers Now

OMG GenAI could cause us to think less! I’m joking but research released today by Microsoft and Carnegie Melon is stealing headlines because it says that GenAI use could cause us to use less critical thinking. Please. AI is not going to make us all dumb. A calculator can multiply 247 times 75 faster than I can, but that doesn’t make me dumb. However, this type of research is worthy to be aware of because when it comes to governing AI, we really do need to explore every risk.

People are angry! Angry that candidates for jobs use AI, like ChatGPT, to respond to questions from employers. For some reason, this old issue has resurfaced pretty strongly this past week. Here is my quick advice to employers and to candidates: just be transparent with each other. Using AI is just another tool so if you let your candidates know your objections to it, i.e. that they may not get the job, then it is on them to take the risk. And vice versa. If you use AI to screen job applicants, let them know so they can try to tailor their resume so they are not screened out for the wrong reasons.

Two megalomaniacs are at it again - Sam Altman and Elon Musk. Elon Musk offering to buy OpenAI for $97.4b and Sam Altman quipped back that he would buy X (formerly Twitter) for $9.74b. It often seems like the feud between those two is made up. I have seen arguments that Musk’s hostile takeover is an attempt to keep AI free and available to everyone. I don’t know about that. I am skeptical of any of his motives but if that was true, wouldn’t he have to be making the same kind of noise at Anthropic or Microsoft? I think the real issue here is that if a non-profit is allowed to easily convert to a public benefit corporation, then why doesn’t every company start as a non-profit?

Wow! New York banned the use of DeepSeek on government owned devices and Congress is considering doing the same for Federal government employees! Okay, for the sake of clarity, we are talking about a ban on the DeepSeek version of ChatGPT. Here is what is true: if you use DeepSeek R-1 chatbot or their API, then their terms of service clearly state that your data will be used in compliance with applicable law. Guess what law is applicable? China’s.

But, it is important to make a distinction. These bans are for the chatbot that you can download from the Apple App store, not the open source version of DeepSeek. Open source means you take the code and run it locally on your own device without calling DeepSeek servers and without sending your data to China. Frankly, this all seems geopolitical. So if you want to talk about it on Valentine’s Day, make sure you have your facts straight.

Caroline’s weekly AI Governance tips:

Is AI governance dead because regulation is out of favor with the Trump administration?

With a Trump Presidency and what happened in France (more on that below), I have seen a few scholars claim that AI governance is dead. Luiza Jarovsky and Sean O hEigeartaigh certainly seem to think so.

Luiza argues that the EU AI Act enforcement just took a back seat to the EU’s desire to enter the AI race. Sean seconds the criticism and adds that the AI Safety Summit draft (from this week’s meeting in France) declaration is woefully insufficient.

With a Trump administration who seems to be executing on their promise of de-regulation and China being relatively quiet recently about AI regulation, I can see why the trend towards AI governance is being questioned.

But I don’t think it is dead.

In the United States, we will see the states get very active on this issue and we will soon be dealing with 50 (more likely 100) laws on AI (one for GenAI and another for other types of AI). The EU AI Act will be enforced because, guess what, it means money the EU desperately needs to catch up in the AI wars.

But you know what I think? I think that of all these nations battling it out, the winner is going to be Singapore. Not only is Singapore a beautiful place to be, but they have been building AI and regulating it very early, since 2019. They have a thriving investment and startup community.

You know what? I just convinced myself. I should move ClearOPS to Singapore!

Chef Maggie Recommends

Chef Maggie is the most popular part of this entire newsletter. Did you know that? Want me to build you a Chef Maggie? I made this recipe a while ago but since it is Valentine’s Day and you may want something fancy and I saw lamb at the grocery store, I will leave you with this marinade for lamb chops - the kids raved about it.

Juice of 1 lemon, 3 garlic cloves, 1 tbsp fresh rosemary, chopped, 1 tbsp fresh thyme, chopped, 1tsp ground coriander (or use fresh cilantro), 1 tsp smoked paprika, ½ tsp cumin, salt, black pepper, olive oil and red pepper flakes if you like spice. Combine altogether and marinate the lamb for 6-8 hours.

Joke: Why did the lamb chef get a promotion? Because they really knew how to rack up the flavors!

AI Bites - 3 Random Things

• If you are interested in Responsible AI, this forum might be interesting to you.

• Are you a lawyer? This is a thought provoking piece on the discoverability of your prompts.

• Still interested in AI Governance? The cohort in February is almost full. Remember your special discount code for being a loyal subscriber “bananas.”

That’s all we have for this week.

P.S. I am a big Morning Brew fan and have been for years and I am so happy they are advertising on my newsletter this week. I listen to their podcast too and I subscribe to three of their newsletters. I highly encourage you to check it out and subscribe!