Ep.24 | The Risks of Giving Your Photos to AI (Guest: Angeli Patel)

Show Transcript

[00:00:00] Hi, and welcome to episode 24 of Privacy Lawls where I Donata Stroink-Skillrud speak with amazing privacy professionals and we have some laughs along the way as well. Today I’ll be speaking with privacy attorney Angeli Patel about the risks of uploading personal photos into AI tools. Angeli is an attorney focused on AI, sustainability and corporate risk.

She’s the executive director of the Berkeley Center for Law and Business. Previously she was an advisor to the White House Office of Management and Budget under the Obama Administration and the US Department of Health and Human Services. So Angeli, thank you so much for joining us. Absolutely. And I still can’t get over the privacy lawls.

It’s indeed very funny. Can you tell us a little bit about the Berkeley’s Innovation and Entrepreneurship Council and your role on a council? Yeah, so, um, like any major bureaucracy, as you [00:01:00] can tell from my background, that’s primarily where I’ve spent most of my time is in large organizations. Um, Berkeley itself is a very, very big place.

It’s known to be one of the most innovative schools on the planet. It puts out the largest number of startups every year, and the country for sure. I, I’d argue in the world as well. And so what, what that takes sometimes is coordinating resources in such a large place like that. It’s nearly impossible.

Every engineering school, law school, business school has founders. They have pots of money where founders can use, researchers can use. There are resources scattered all throughout the school. And so the role of the Innovation Entrepreneurship Council is to really aggregate and find a way to connect students and researchers.

To innovate whether that’s, uh, that they need legal help in incorporating, whether that’s they need business help and taking an idea to market. Any number of, if they need space, if they need mentors. [00:02:00] The INE Council essentially helps. Anyone who’s trying to innovate on campus, it’s, it’s for students and researchers at Berkeley.

So my role is representing the law school and, uh, at Berkeley Law, you know, we’re one of the largest schools for tech and privacy. And a lot of our students also focus on. Becoming lawyers who support the startup ecosystem. And so, uh, the law school also plays that role on campus for Berkeley innovators as well.

And so I kind of. Play the multi hat role of both being a resource myself the Berkeley Center for Law and Business being a resource for, um, startup founders, but also the larger law school. That’s awesome. That’s really cool. Um, I wish I had that when I was in school ’cause there was nothing similar.

And then when I decided to start this business, it was like, okay, well do you go to score? Do you try to like find different [00:03:00] resources online and. You know, it’s hard. It’s hard to know who to trust or what’s the right information. Yeah. So having that right at your school must be really nice. Well, it sounds, it’s, it’s very nice, but it’s also almost necessary just given the size of the school itself.

And I think, you know it’s, it’s not a perfect solution, but it’s definitely one of the, one of the better ones. What it does also is it helps a lot of lawyers actually start to think innovatively. Mm-hmm. So a lot of law students themselves want to join the world of entrepreneurship and. Being kind of plugged into the larger campus community where there aren’t just lawyers everywhere.

Yeah. It really, really helps. It’s very, very, um, what do you call it, cross pollinating, that’s the word. That’s awesome. Cool. You recently wrote an article about protecting the rule of law as attorneys. Can you tell us a little bit more about, you know, why doing that is so important? Yeah. As you can tell, I am, I am working on a lot of different topic areas as part of [00:04:00] my, both my role, but also my personal interest.

And, you know, more recently with all the things that we’re seeing in the news around judicial officer safety to generally, any third party, if you would, into how the, the law is practiced or how justice is administered. At least for me, it’s, you know, I think we should all kind of take a moment to think about it and reflect on what that means to us as individual attorneys. Uh, I, I think we live in a time where our institutions are.

For better or worse kind of going through it. Mm-hmm. And it’s a time for a lot of attorneys as folks who interact with legal institutions and are in some ways part of those institutions to really think about your own individual role in, in what you believe in, in that institution, or whether that’s the rule of law generally.

I think for me what that means is, and the reason why I wrote that article was like it, it was in order to kind of nudge us all, [00:05:00] identify what our moral North Star really is. Individually, I think there’s a lot of organizations we are part of a lot of bar committees, a lot of political parties, a lot of schools.

And you know, there’s so many ways to identify as a person, but ultimately you’re the one who’s practicing and you’re the one who has to make judgment calls. And I think coming up with your own version of what. It means to protect the rule of law. What it means to be in a democratic society for you is something that we should all take time to, like write down somewhere on a piece of paper and say like, here are my red lines, here are my boundaries, here are my commitments.

And so that’s kind of why I wrote it. For me, what that means is, you know, it’s always meant that the you. Using law in a way that you know, continues to create. And at least I’m a corporate lawyer, I’m a privacy attorney. I work on tech and tech matters, so I don’t have the opportunity to go in front of a judge as often as maybe a lot of other our litigating counterparts might.

But for me, what that looks [00:06:00] like is ensuring that you know, where there are loopholes or there are gray areas. We tend trend in the direction of accountability, transparency whether that’s in governance, whether that’s in contracts and transactions, you know, making sure every party there’s good faith where there is good faith.

I think there’s a sport a lot of times to being a business attorney and it’s really fun. But I think that there are lots of different places where we can practice what our our, our moral, boundaries are. So, you know, the article really is just about do your reflection, figure out what it means to you in your practice what it means to be an attorney, that that values the rule of law.

It’s interesting because so many attorneys who, like myself, I don’t go in front of a court and court, and I don’t practice traditionally either. Um, so for me, like I’ve been a part of the American Bar Association’s rule of law initiative for many years now, and initially it was focused on the rule of law abroad, and it seems like it’s [00:07:00] focusing a lot more on domestic rule of law.

So being part of those groups as well is, is a big way to further that. Um, you know, even if you’re not going in front of judges or you’re not Yeah. Practicing traditionally either. I’d even say for, for so many you know, corporate attorneys, specifically, if you’re advising companies, I think there’s a lot of really great ways.

I think a lot of, a big part of modern corporate practice is, you know. Encouraging companies to have a more proactive stance on regulation. Mm-hmm. Or having more proactive stances with engaging with government. And I think those are the areas where you can be like, alright, this is, this is where the rubber meets the road.

And we can start developing like best practices and, you know, the rules of the game that don’t skew public and government power in a way that. That it, you know, unfairly it’s, it’s benefiting maybe one company or one industry. So I think those, there are like small places where that can come into [00:08:00] practice, but yeah, I agree.

It’s, it’s absolutely, it’s rare. So we’ll talk about the risks, the privacy risks of uploading your own photos into ai. But before we get into that, I was just kind of curious. Do you use AI in your personal life, like outside of work and outside of your community work that you do, but just in your personal life and how do you use it in your personal life?

Oh my goodness. I use it probably more than most people, so I use it in a number of places. I have my journaling AI that I use, which is called Mandela. It lives in Slack, and it’s basically kind of is like a personal coach that walks me through like, I’m having a bad day, I’m really mad at this person.

What do I do? And it like coaches me through difficult situations. So I have like my mental health leadership coach. I have my, uh, multiple use case for Chad g Bt on here’s what’s in my fridge, what should I cook two. You know, I see something walking on the road and I wanna know what it [00:09:00] is to any number of those kinds of interactions with chat GPT.

I absolutely love all of my note taker and my voice recording ais so I could talk to it and it can transcribe my thoughts, um, and gimme insights. So yes, I do use AI quite a bit. I also really enjoy using the AI features that existing kind of things that I use have, like Canva, for example.

I use Canva a lot for designing a number of things and their AI feature to generate a new image you might need or. Uh, a new way to configure a template. So I know it’s kind of everywhere now that, now that you’ve asked this question, I’m like, where else do I use this? I also use it in my Aura ring. So I have wearables and I, I’m clearly very tracked and measured person.

But Aura has like an AI advisor and it gives you insight into your health. So yes is the answer. I use it a lot. I haven’t even thought of like half of those uses. ’cause usually I’ll do, um, [00:10:00] so I’m Lithuanian, so if I need to like translate a certain, like more complicated concept into Lithuanian, I might use AI for that.

Um, or I may use it to like write notes to family members. I’m like greeting cards, but I can’t think of what to say apart from like happy birthday, but that’s really it. I did not. Using it to like plan your meals is really smart because they always have stuff in in the fridge and I’m like, what am I supposed to do with this?

And then I look up recipes online and they have like 30 other ingredients that I don’t have. I’m just like, okay, this won’t work. And that’s really smart. I haven’t thought of that at all. Or if you’re a plant mom like myself, then taking pictures of your plants and asking what’s wrong with it is a very, very helpful tool.

I just assumed that my plants don’t have what it takes to survive in this fast-paced environment, so I just not cut out for it. Exactly. Yeah. Half of them are teetering between life and death at like any given moment. Lovely. [00:11:00] So. We are talking about, uh, we’re gonna talk about uploading photos into ai and I think that, you know, it’s like human nature to be curious and wanna try new things and participate in what other people are doing.

You know, we used to have the Tamagotchis or we had, you know, you wanted to go with prom with your friends, or you wanted to make friendship bracelets and you know, now we’re experiencing AI and there’s so many AI trends that. People wanna be a part of just because it’s fun. But those trends can introduce some serious privacy risks and we’ll talk about that today.

Um, and we’re keeping our focus really narrow. So the privacy risks of uploading your photos to an ai. Can you tell us about some recent trends where people are doing this? Like the action Doll trend or the Studio Ghibli trend? I think there are a lot of trends, um, and I think they’re very prevalent in I think the younger generation.

So Gen [00:12:00] Z and Gen Alpha, though I don’t think that they’re the only ones. I think millennials and anyone who’s using Instagram or TikTok is probably at least tempted to participate in some of these trends. I think one that I might even point out is the. I forget what the name of the trend was, but what your child will look like.

You could come up with like, what you think your future child might look like, that, that trend, whatever it is. I think there’s an action figure trend. You know, there are a lot the starter pack trend. There are a lot of different trends that are very popular. And I think, you know, this. Is a very, very popular tool for collecting data.

I don’t think when you see a trend you’re thinking, oh, what a great data collection method. Um, usually you’re like, I wanna know what my studio people like portrait will look like. Yeah. And it’s completely valid. It’s, you know, it’s very, like you said, human nature to be curious [00:13:00] and wanna be part of something.

And. The trends that make it are the ones that do in fact tap into the nostalgia or into, you know, a truly kind of out of the world creativity experience. And. I think that what users when they’re uploading their photos may not be thinking about is, you know, just how drastic that one move is. You think you are getting a really fun output and that it, that it ends there, but it is just the beginning of that large and long and permanent and potentially forever evolving profile of you that lives somewhere.

Many somewheres in the ether and in, in certain databases that are for-profit, nonprofit, dark market, black market, wherever you, your data may end up and it is so hard to track at some point. I think it’s also important to realize that [00:14:00] a lot of these companies have absolutely no incentive to.

Protect your privacy if they can disclose it upfront to you and otherwise be fine. Meaning if you’re not one of those people who’s sitting there reading privacy policies for a living or just because you like to do that. And I know for a fact that most people don’t, and I think this is the lifelong, and since its origination issue with privacy policies is no one wants to read them.

No matter how easy you make it, no matter features you add to it no matter how quick and easy the click through process is. It’s just. Onerous for individual users to read it, especially when a lot of the AI apps are embedded in Instagram or other third party applications. Mm-hmm. Um, and so the [00:15:00] privacy implications of uploading your picture can truly just be unknown.

Um, on top of that. And I can, I’ll I, tell me when to stop because I can keep talking about how bad this is. And I’m also, you know, caveat that by saying like, it’s completely understandable that someone would want to upload their photo, right? Like it’s, it’s one of those things that at some point all of us are users of social media and already we are putting so much of our personal data and personal imagery out there.

I think that, you know, even with privacy policies, you know, let’s say you were to read it and let’s say you were to understand it and let’s say it didn’t conflict with itself. ’cause I’ve seen so many policies that start off with, we don’t sell your data and we would never do that. And then you scroll down and they say that they do, and here’s the companies they sell to.

You know, those policies can change or the company can violate their own policy. I mean, there’s so many cases out there and so many enforcement actions where the company [00:16:00] said, we are going to do this with your data, and it turns out that’s not what they’re going to do, or that’s not how they’re gonna share it, and that’s not how they’re gonna sell it.

So it’s not necessarily, you know, you have to have a trustworthy company behind that policy in addition to having the policy in place too. So. You know, let’s say I’m, I’m curious. I wanna try this out. I’m gonna upload my photo into an ai Does, do the AI tools run like biometrics on your face? Yeah. Yeah. So AI tools are able to, you know, using, and I’m not the AI technical expert, but from a baseline it can use whatever it’s neural network.

Underlying are to actually run facial geometry, run other kind of analyses from the pictures that you upload, and essentially collect biometric information to identify what it’s looking at. Um, [00:17:00] whether that’s a fingerprint or a face print or any number of those things. It’s the same technology that your iPhone uses or what TSA pre uses.

When you go to the airport. But now it’s so much more accessible and so much more anyone can access and upload using Chad, GPT, any number of these AI tools. So yeah. Yeah. It, it in fact does run biometrics on your face. Yeah. And that’s kind of what you want it to do, right? When you’re partaking and these trends, you want it to be like, well, what do you see?

Like, what do I look like? You know, am I, what is my, like, you know, twin in like another universe look like, you know. Yeah, you want it to be an accurate representation on your, of your face. Like you don’t necessarily wanna upload a photo and then get an image that’s like of somebody completely different.

But that comes with a lot of privacy risks, right? Like biometrics, you can’t, I mean, you can change your email, you can change your phone number. It’s a pain in the butt, but you can also change your social security number, but you can’t change [00:18:00] your face. And that’s the scary part about this is like once your biometrics are out there, that can be used for all different kinds of reasons.

Like, you know, unlocking your phone, falsely identifying you or identifying you at different places, like a protest or, you know, with women at a women’s health clinic you know, all those types of things. And I think that’s, I think that’s the understated part of privacy risks. I think because we live in such a digitally connected world, there’s almost like a throw your hands up, we can’t control anything.

What do we do anyway? Kind of hopelessness even around privacy. Um, responsibility from an individual standpoint, but I think it’s really important to know that there are in fact companies that are, you know, unicorns that get insane amounts of valuations and funding to collect and aggregate, you know, various data points from various providers to build user [00:19:00] profiles, to build human profiles and to, you know, be able to identify you from publicly, you know, recorded.

You know, cameras on the street to things you upload, to places where you’re in the background somewhere on somebody else’s photo, right? So, uh, it’s important that there’s such a huge market incentive to build really strong. User profiles. There’s so much money to be made in that business. There is, there are so many customers, and they’re not individual customers.

They’re governments, they’re mass marketing agencies. They’re. You know, without naming any individual companies, but anyone who’s stands to benefit from ad revenue. Yeah, there’s a lot of, there’s so much incentive. And it’s, that is why it’s so important to be aware of uploading your picture because biometric data is uniquely one of those things that [00:20:00] you can’t really change it.

If you get a nose job, maybe that changes it a little bit, but even then, you really can’t change your face significantly enough. That would make you on a unidentifiable. It, it lives on you, it’s immutable, and it’s for that reason. You have so little control. Then you think when you’re uploading. And you’re just, you know, kind of giving this up for free, right?

You’re not even, I mean, yeah, you’re getting like a fun photo in exchange, but that’s not really worth it when you think about all these companies that could abuse it. It could end up on the dark web, it could be used to steal your identity, it could be used by authoritarian governments. Like all of these things that could happen with this face print.

And even if right now you don’t necessarily care, like there’s nothing that you’re doing that could put you at risk. You know, let’s say 10 years from now, you join a movement or you join a protest or, or anything really, and all of a sudden it’s a huge problem for [00:21:00] you and you’re like, wait, how did they get this print?

Well, it turns out you want it to look to see what you look like as a baby 10 years ago, a thousand baby pictures of yourself already. You know, it’s just, yeah, sometimes it’s just not, not worth it. I think, I think you’re right. I think being able to, you know, not get stopped at the airport, I think, you know, a lot of, you know, a large part of the population who are primarily Brown experiences this during nine 11 is, you know, that facial profiling, uh, when you’re crossing through security, can you imagine.

Even having a lot less predictability when you go to the airport or any public place where there is security for who knows what reasons, right? And so, or in other countries, you know, here you are getting a visa, very benignly, and then now you’re on some sort of list. For what it hard to say, ’cause there’s no procedure, there’s no due process that helps you walk through those things.

[00:22:00] Um, so the risks are real. And this is not to say we should all live in fear and, you know, but it is a, a call to be very, very responsible for who you’re giving your photos to and whether that immediate gratification of participating a trend is truly worth it. That’s, but, but that point you made about.

Not getting paid for it, like it’s like mm-hmm. Giving it up for free. I think that’s, and like, I don’t think you can truly ever get paid enough for things like that. Um, but like, uh, I, I’m thinking about this. I dunno if it was like a case or I read somewhere. There was a company that was sued by a class action of consumers suing under a privacy, I think it was like in Indiana, I can’t remember, for biometric data.

And the judge granted all of the. Class members equity over the company. Hmm, interesting. Yes, and I think, I guess why would you [00:23:00] want equity in a company that just violated your privacy rights? I’ve seen those before where there’s like class action lawsuits, multi-billion dollar settlement, and then each class member gets like 25 bucks and it’s like, what am I supposed to do with this?

I’ve definitely been part of those and I’ve cashed many a checks for like 10 bucks. But it’s like, you know, your privacy was violated. All your data ended up on the dark web. You had to switch all your credit cards, you had to change all these accounts, you had to change all your passwords, and then at the end of the day, get $25 for it.

And it’s like. I mean, who is this for? Right? Like was this really compensation of the amount of time that it takes consumers to like perform all these actions because of the data breach that you people caused? I don’t think so. I don’t think so. I mean, a hundred percent agree. Although if it does dilute the equity of a company, I do think at least out here in Silicon Valley, the most coveted.

Currency is equity and if [00:24:00] it, I think it was like some, like 30% of the company’s equity went to consumers. Wow. That’s pretty significant. I’d, I’d say. Yeah. That’s, that’s not, not a bad deal. Yeah, it’s not bad deal. So I, it’s per consumer. Maybe it’s not like significant, but if I forget how many people were in that class, but.

I say for the company, it’s crippling. Right? Yeah. And I think perhaps that was kind of the goal there is to a bit of a punishment in disguise. That’s a good one. Yeah, that’s a new one for sure. You don’t see that. Um, I think the company was called Clearview ai. Clearview ai. Oh, interesting. Okay. We know. We know Worth looking into.

Yeah, you know, those people. Um, so in addition to getting like biometrics and things like that would the AI also receive metadata and, and what does that metadata expose? Yeah. Yes, indeed. Lots and lots of metadata, I think. Pull up any picture on your iPhone. I’m assuming majority of the listeners are iPhone users.[00:25:00]

Any Android phone as well. You kind of look at the info button on your photo. You start to see time, date, location, all this other stuff that I don’t understand. You, you can leave notes on that photo. You can share and give access to other people on your phone to that photo. There are, it can tell you, um, uh.

If your photo is taken in a live mode, it can have a number of photos within that one photo file. And so each photo has a lot of metadata and I think you can turn those features off on your phone or whatever device you’re using. But it’s really important because metadata is basically other information attached to your picture.

Outside of the image itself, that gives away additional information about who you are. This could also include what kind of phone you used your, your address your GPS coordinates, you [00:26:00] know? And this just continues building your user profile out there somewhere along with what you look like.

So now, not only do they know what you look like, but they know what kind of phone you have. They know what time of the day you tend to take these photos. They know where you are when you’re taking them. And that is a lot of information to be uploading. For a trend. That’s, yeah. Yeah. I mean, you could, yeah.

I feel like you could also build like quite a profile about a person if they upload a lot of pictures, you know, let’s say they upload like multiple pictures from their workplace, you know, where they work or maybe upload pictures from their church. So. You know, their religion. The thing that really, um, scares me about that, the metadata is the idea that like I’m just over there in the street and I take some picture.

Somebody obtains that metadata and there’s like a crime committed in the area, and all of a sudden I’m like caught up in some [00:27:00] criminal investigation, even though I had nothing what to do with it. But the metadata showed that I was in like similar vicinity around that time. Yeah, that part really freaks me out.

I, yeah, it’s very freaky. I also think it’s really scary to be able to build a profile about you and your associations. Like mm-hmm. You all you did was upload full picture pictures. Maybe you’re one of those very cautious folks who doesn’t put like, too much information about what schools you went to or where you work on your, on the internet, because you don’t wanna be identified in any way.

But then you have a share picture and you just happen to be on campus. So, or you happen to be coming away from work over and over, and a number of your photos show you know, what your associations are At this point. Now you, now you have like such a more, there’s a richer version of you that lives on the internet for someone to understand and, and the use cases you can get creative with, right?

You can get very, very creative at the most benign level. I’d say maybe now like, um. I don’t know. You know, [00:28:00] a consumer product company’s like, well, we really, really wanna sell, I don’t know, these really cool, like self heating water bottles to people who live in cold climates who, you know, have a big commute.

Great. I’m sure they can figure out that X, Y, Z person tends to take a picture from this bridge a lot. While they’re on a drive potentially. So they must have it right. And that’s benign right now. They’re gonna target you with this. Cool. For a, but it could get worse, right as you said it. If you took a picture at a church or some sort of group and you happen to be in a jurisdiction where certain protections are not afforded to folks like yourself, um, now you’re in real serious trouble.

So, yeah, no, the risks are real. Absolutely. Or maybe you are in a jurisdiction that does afford those protections, but then you travel to one that doesn’t, or you wanna move to one that, you know, just happens not to. To offer that. I mean, [00:29:00] I’ve also heard of people, you know, this data could end up on the dark web and could be used to make deep fakes or pornographic images of you with your face on, with zero consent or zero knowledge from you.

You know, all of a sudden your faces on all these websites doing things that you yourself would never do, or at least not do in public, and there’s not much you can do about that. And I think, you know, I think we misunderstand the severity of that, what you just mentioned, the deep fakes. Mm-hmm. And the potential, because you know, I think there are a lot of great companies that are coming out that are working on exactly identifying fake images.

Images fake, you know, there is an interest and market incentives, if you would, for that kind of data and for that kind of technology to exist. And yet once, once something like that’s linked to you, whether or not it ultimately turns out to be fake or real or anything like that, [00:30:00] there’s an emotional toll that you’re dealing with.

There’s an emotional and social toll that you’re gonna be dealing with, and I think we underestimate. Because we, you know, we’re lawyers. This is a, a legal conversation to some degree, but I think there’s a human element here that far goes on beyond the legal repercussions that something can have including reputational.

So when you do apply for a job and all they can find is a deep fake of and pornographic video or pictures. Or perhaps very compromising, politically sensitive things. Mm-hmm. Um, you’re, that’s a very tough one to walk back from. Yeah. And you know, imagine explaining that to your family and your grandparents or your parents who may be older and don’t understand this technology.

Like, Hey, this is fake, I swear, you know? And. Yeah, there’s no better way for people to think it’s real than to say it’s fake. And, and that can be really, really bad. It’s again, in exchange for a picture of [00:31:00] seeing what you look like as a baby or seeing what your twin would look like or whatever, you know, it’s just, yeah, sometimes it might not be worth it.

You did mention previously the, the privacy policies of these AI companies, and we talked about, you know, how the company itself needs to be trustworthy for you to be able to trust the privacy policy. But some of these like more famous AI tools you know, do they say that they’re gonna sell this data?

Do they say they’re gonna share this data? Like what do their policies say about this? Yeah, I think a lot of companies. Say that they’re not going to sell this data unless you opt into that. So there’s a lot of opt in abilities. I think that there’s a lot of companies that also, I think, I’m trying to remember what Open AI’s policy versus like anthropic.

Those are the what, the two main ones we perplexity. But each of these companies do not make it clear and are not out there to say we [00:32:00] won’t. This is a very case by case thing. Um, certain tools and certain, um, GPTs for example, on Chad GPT may expose you to your data being sold and trained upon.

Other GPTs won’t, and so there are, there are a lot of carve outs within privacy policies themselves that, um, you do need to be very careful and read very line by line as to what they’re promising. A lot of the privacy policies, forget resale. I think that’s, you know, I think sale of privacy data is a, is very concerning and that’s a very obvious, like kind of low hanging fruit to identify as a risk.

Mm-hmm. But I think that there are kind of certain embedded risks that perhaps you might, as a user not be really thinking about, which is storage. And kinda the reuse of certain data and then of course training those models. And I think the training those models piece for me is what really sticks out as equivalent to [00:33:00] selling my data, uh, largely because now there’s in no way can you remove that.

Data ever again, right? It’s now within the LLM, it’s being trained upon other data is being created using my personal data. And that’s something that those companies don’t promise that they won’t do. The major companies, um, there’s uh, I think Chad g, pt, or sorry, OpenAI says that you can opt out. Or opt in, I can’t remember exactly the, but there is a feature where you are able to say, don’t use my data to train your lms.

Uh, but the whole reason why GPT and Claude and all these things are available for free is for them to collect data. Yeah. And then for you to pay of course for a more premium features, but it’s really about getting enough information. Yeah. To me, like training the AI is almost like your data is now part of the public domain.

That’s how I look at it. Yeah. Yeah. And it, and like for better or worse, right? You can argue that it shouldn’t be that [00:34:00] way. I agree. But it is, it’s something that, it’s just nearly impossible to extract back out once it’s out there. Mm-hmm. Um, and you know, kind of reminds me of the right to be forgotten and the.

European Unions kind of attempt at being able to delete your personal, I just don’t know how that’s gonna look now, especially if certain data is being trained upon. So you remove, you take down a picture from Facebook, but what if an LLM has already been trained on it? Then what? Right. And those are the kinds of the questions that haven’t been answered yet, which, you know, makes it up for, it’s, it creates risk.

I think I saw there were a couple cases where, um, they made the, the AI company delete the entire training data set because there are some like PII in it that somebody wanted removed and they didn’t remove it and the person filed a complaint and they’re just like, okay, you have to delete your entire training set now.

Um, and there’s, yeah, but, [00:35:00] um. It, it’s pretty wild. I mean, there, there’s also instances where, you know, you go into open AI and it tells you you’re working with an ai, but it doesn’t necessarily explain what’s happening or if your data is part of a training set or what that means. There’s also, you know, instances with like the Facebook ai.

Where people were interacting with it. They got a prompt saying, all the stuff you’re gonna put into this AI is gonna go public, like it’s gonna be posted on this public form. And they continued proceeding and then asked questions that you wouldn’t ask if you understood that this information was gonna be become public, like how to cheat on a test or like.

You know, uh, very, very private things. And I think that even when that notice is in place, a lot of consumers, you see that box are so used to clicking okay or clicking X out of those boxes. They don’t necessarily even read it. So, I mean, even if you’re trying to tell people it’s gonna become public, sometimes that doesn’t even work.

So it’s really tough problem with how to [00:36:00] make people. Give them the right information, make them pay attention versus, you know, not say anything and just collect this data which you’re financially incentivized to do that. Well, and I think that what you mentioned around, um, you know, consumers kind of glazing their eyes over mm-hmm.

Any given, privacy notice or risk notice. It’s in some ways it’s, it’s exactly what you’d want to happen if you were a company. Yeah. And this is not to, not to malign any companies because a lot of, a lot of startup founders, for example, that I work with, really don’t want to fall into the privacy issue.

They really, really want to do the right thing, um, when they’re building their AI products. But it is one of those places where you’re like, how do we get people to really understand what they’re doing here? Yeah. Like how do you tell someone like, please don’t tell us how old you are, like what color your [00:37:00] eyes are.

I don’t wanna know the nickname members of your family. Like, yeah, but like, how can you, how can you, you know, don’t gimme your address like. It, none of these things are important to the product you’re about to use, and yet people wanna share and they wanna access the richness of certain products. So it is very difficult.

And the other thing that kind of, uh, you know, what you were saying. Around kind of the risks that these tools tend to run and, and having to delete entire training sets. I think that’s fantastic because that risk is gonna create, as you know, my, my job, my day job, my job is to kind of think about like, what can we predict about the market around AI for the next five, 10 years.

You are going to see, I mean, if, if it is true that it takes so many resources, energy, and data collection in order to build really strong and accurate lms, and then by contamination of having PII. [00:38:00] You have to now by law, because courts demand it have to delete. There’s even a risk that you might have to delete all of that.

Yeah. It’s so costly that now there is an incentive to ensure, at least by those companies who don’t wanna face this outcome to make sure that PII is truly, truly scrubbed. And if it is not doable right now, then there’s a huge opportunity for innovation to be creating data sets that can be truly scrubbed.

So I think there’s a little bit of. Hope. I think given just how powerful AI is as a technology to me, perhaps build in a way that you can actually achieve total privacy. Maybe this is me being hopeful. Well, I hope so too. That’d be nice. Um, so we could use more of these tools ’cause they are, you know, extremely helpful.

Um, what tips would you give to consumers who want to use AI but who also wanna protect their privacy? Yeah, and I think, I’m gonna sound [00:39:00] like beating an old horse dad, but there’s, you know, pause before you upload anything. Make sure, especially if it’s a photo, make sure that the data metadata is scrubbed.

You can do that on a number of ways. That’s, that’s the first point. Just taking a moment before and just thinking about what you’re doing. I think there’s a mindfulness element to using AI tools and honestly using social media generally, uh, because those tools make it so easy to upload and share and engage.

Then that brings me to my second point is to actually understand the platform that you’re using. If you’re doing this likely on social media, understand that the incentive here. The things that are pushing you to do what you’re about to do are truly in order to get your data, there is no other incentive.

It is not about building community. It is not about building connection. The only goal of a trend is to collect data, and I wouldn’t have [00:40:00] said that before, but I say that now because it is the only incentive of a tool like Instagram or TikTok is usage engagement collection. Yeah, it’s, it’s funny because, um, so many of those tools, you know, their ads or whatever, or posters, you know, let’s build a community, let’s share our interests and we’re all interested in knitting and so fun and we’re all gonna get together and.

But in reality, you don’t need Instagram for, for doing any of that. You know, you can meet with your friends in your knitting circle or whatever. Like, it doesn’t have to be like this. It doesn’t have to mean that you’re giving up all this data in exchange. And I think that’s why I point out the social media component over what I just mentioned.

Like I use Slack and they have AI tools. It, the difference here is that social media is kind of like the. Engine on steroids designed [00:41:00] to collect data. That is a known incentive of social media companies. That is literally what they exist for. Mm-hmm. And so if that’s where you’re accessing AI tools and that is likely where you are using biometric related AI tools, um, that’s where I say pause very much breathe.

And take a moment before you upload. Yeah. Additional information. Do you have any like favorite resources for privacy attorneys to learn more about ai? Because I know we’re getting more and more of those questions every day. You know, I could say there are podcasts and all these things, and yes, there are, but one of the places that I found very one of the books that I read that I found really, uh, illuminating was the Alignment Problem.

And that’s an incredible book. I recommend it to anyone who’s trying to understand both the technology and kind of the issues with the [00:42:00] technology in a way that’s both constructive and understandable. The book is really good at helping you understand what the backend of AI really is and how it came to be.

And I think as attorneys we shy away so much from technical and I’m no different. I get it. Same. That’s why I joined the law rather than being an engineer. But I think it’s, if there was ever a time where we should know how certain technologies work, whether or not you practice tech, law, privacy, business, or anything.

It’s now because now this implicates all of us. And it, it also is, is going to be used in reasoning, legal reasoning, judicial reasoning, um, prediction. Um, so it’s, I, I just could not recommend that book more. It’s called the Alignment Problem and if I, I cannot remember the author, but the title speaks for itself.

Awesome. Thank you. Um, any final thoughts before we kind of conclude here? Yeah, I mean, final thoughts [00:43:00] are, you know, I, one, I just absolutely love that these are questions that are being asked right now. I think I have a lot of hope. I always end most of my conversations on a hopeful note, uh, which is usually rare on, on, on AI related topics.

But I have a lot of hope that as much as there are risks to privacy in the world of ai, I think there’s a lot of money. Market and need for privacy, innovation and this, there could not be a better tool and technology to help us achieve that than ai. So I feel very, very hopeful about that. Awesome.

Yeah, I, I hope so too. I hope we can get some legislation. I hope we can get some rules. I hope that companies follow those rules and, and if they don’t, I hope that people are, um, playing it safe with these tools too. Or just anticipating being part owners of majority of the AI company. Well, thank you so much for joining me today.

I really appreciate the conversation. Um, and for [00:44:00] anybody listening, um, definitely make sure to subscribe to Privacy Laws, um, so you don’t miss our next episode.