Taylor Lorenz’s Power User - AI Deepfake Panic is Killing Free Speech

Episode Date: July 18, 2025

There's a bill right now in congress  called the NO FAKES act that claims it will protect you from deepfakes. "In this new era of AI, we need real laws to protect real people,"  Rep. Maria Salazar said when announcing the bill. "The NO FAKES Act is simple and sacred: you own your identity, not Big Tech, not scammers, not algorithms." As usual, none of this is what the law actually does!!!!! The NO FAKES act is actually an incredibly sloppy and dangerous piece of legislation that could destroy lives, eradicate privacy, and lead to sweeping censorship of journalistic and constitutionally protected speech. There's a reason that major civil liberties orgs are sounding the alarm about this proposed law. The NO FAKES act is "something that could change the internet forever, harming speech and innovation from here on out," The Electronic Frontier Foundation recently declared. This bill would be a disaster for online speech and free expression. The Center for Democracy and Technology published an open letter opposing the proposed law. Kate Ruane is the director for the Center for Democracy and Technology's Free Expression Project. She joined this week's episode of my Free Speech Friday series to break down what the NO FAKES act really says and why we need to fight back against it.***** Buy a subscription to my Tech and Online Culture newsletter, User Magazine to support my work!!!! 🙏 https://www.usermag.co ***** Subscribe to my newsletter: ⁠⁠https://www.usermag.co⁠⁠https://www.instagram.com/taylorlorenz https://www.instagram.com/taylorlorenz3.0 https://www.tiktok.com/@taylorlorenz

The number of pictures and videos threatened by those filters was actually tiny in comparison to what no fakes is going to do. There's a bill right now in Congress that if you didn't know any better, sounds great. It's called the No Fakes Act, and it claims to protect you from deep fakes on the web. Representative Maria Salazar from Florida said this when announcing the bill. Quote, in this new era of AI, we need real laws to protect real people. The No Fakes Act is simple and sacred. You own your identity. Not big tech, not scammers, not algorithms.

But as usual, none of this is what the law actually does, nor is it what the law is actually really about. The No Fakes Act is actually an incredibly sloppy and dangerous piece of legislation that could destroy lives, eradicate people's privacy, and lead to sweeping censorship of journalistic and constitutionally protected speech. There's a reason that major civil liberties orgs are already sounding the alarm about this proposed law. Kate Rune is the director for the Center for Democracy and Technology's Free Expression Project. She's joining me today to break down what the proposed law actually says and why we need to

fight back against it. Hi, Kate, welcome. Hi, Taylor. Thanks for having me. I'm so glad to be here. So to start off, can you explain what the No Fakes Act is and what it does? So what the bill says is if somebody out there in the world uses your image or your voice to create a piece of media, be it audio, video, or just like a static image, they cannot do that without asking your permission first. So they have to get a license from you or whoever owns that right to create that piece of media or that piece of content and then distribute it. I feel like if you told most people about this, they would think that that sounds like a good thing because we all want control over our images. And I feel like we're all hearing so much

about these deep fakes and people's voices or images used without their consent for nefarious purposes. So can you kind of explain why this bill might not be what it seems? Like, will this affect everyone or just famous people? The way the bill is constructed, it's supposed to be a protection for everyone. But it's mostly going to wind up as far as I can tell benefiting public figures, famous people, and gigantic rights holders like record companies and movie studios who are going to be incentivized to get who signs contracts with them to license their brand new right to their face or voice to the company.

And then the company will be allowed to go and defend that right or use those images and voices as they see that going forward. So it looks like it's a protection for everybody, but what it really winds up being is a method to suppress dissent on the part of powerful people like politicians and public figures who don't like it when people criticize them or make fun of them. also going to have like some really interesting impacts on historians and academics who are going to use this technology to try to illustrate things like historic events because not only does this give a right to everybody in their image and likeness, it also gives this right to dead people. So dead people's errors and assigns can assert this right on behalf of people who are dead. And if you do things that they don't like, there will be a method to get that speech removed from the internet.

I want to back up to what you just said about how this will sort of allow powerful people and politicians to potentially censor criticism or parody. How would that work? The way the law works is somebody creates a digital replica and places it on the internet. Under the bill, online service providers are required to allow rights holders to send notice to them if there is a digital replica that is up. And then they are required when they receive that notice to take it down. They are required to confirm whether it is a digital replica, but they are not required to confirm whether it falls into any of the existing exceptions to what a digital replica is. So if it's in a news story, if it is an academic or scholarly work, if it is parody or satire, the online service doesn't have to look into that. And because they would be on the hook for lots and lots of money, penalties of up to $750,000 in statutory damages, plus more on top of that, if they leave content up, they're really incentivized to take it down.

You also mentioned that this law could affect historians and academics. How would that work? So let's imagine, for example, a college professor who uses generative AI to recreate a documented conversation between historic figures where one of the figures expressed racist beliefs in support of continued racial segregation of schools and public spaces. And the professor posts this lecture, including the AI generated conversation on YouTube. Now, I think we can easily imagine that heirs of this segregationist. might not like being reminded of this truth about their history, and they request to be taken down.

It is a digital replica. It's an unauthorized computer-generated image in which the depicted person did not actually appear. It should fall into an exception. But again, there is no requirement that anybody, not the rights holder and not the online service provider, check for that. So the odds are that the lecture depicting a conversation that has been historically recorded would be taken down. As a real-world example of that, Daphne Keller, who is a professor over at Stanford, she has a fourth-com piece on the No Fakes Act and in it, she notes that she used altered video, the altered video of Nancy Pelosi, where the former speaker was made to look like she was slurring her words as part of an exam question. If Daphne were to post that exam question online and include the video in it, which Daphne often does things like that,

that content could be subject to take down under the No Fakes Act should Nancy Pelosi decide that she doesn't want it up there anymore. So that's exactly how this is going to work, the incentive structure the bill creates, is strongly on the side of removing content. And there are very few enforceable protections for free expression. As a journalist who reports on AI and deep fakes and things, I feel like often media organizations need to kind of show or dissect some of these deep fake videos to actually debunk them. Would that sort of journalistic work also come under attack under this bill?

Almost certainly. I actually think that journalists in this space, that this bill purports to be trying to address are potentially at the most risk here, because one of the most useful journalistic things that can be done here is for journalists to do exactly what you just said. Take the newsworthy, deep faked content and present it again in a news context for people to be able to digest and understand better so that they can see how good the content is and how realistic it is, and then understand the ways that they might be able to identify it going forward. But under the No Fakes Act, anybody depicted in that content can submit a takedown request.

Just because it's in a newsworthy document, again, there's no requirement that anybody review for an exception applying. It is incredibly likely for that content to come down. And then another thing to know about that particular circumstance is it is then incumbent upon the user, upon the creator, so upon the journalist, to file a lawsuit in order to keep that content up or in order to restore it, which is incredibly burdensome. And essentially, one, not how First Amendment litigation works in any other context. It's usually the person who is accusing someone of saying something or doing something illegal speech-wise who has to go to court. And two, it's also not how the Digital Millennium Copyright Act works, wherein in that circumstance, the user can provide counter notice and then it is the rights holder

who has to bring a lawsuit so that there's a significant burden shifting that is happening under no fakes that we haven't seen before. Yeah, because obviously I feel like the people that are the journalists, the parody artists, like the people online that are kind of using technology to like debunk things or speak truth to power or make parodies, like they're not going to have the resources to fight back against any of this. It'll just get taken down. That is a huge concern. And I also want to point out that this is also going to be dangerous for real content. So the bill pays lip service to real content. It says that real content isn't

supposed to come down. But the reality is there's very few incentives for rights holders not to submit questionable notices. And there is basically almost no way for online services to tell the difference between real and fake content in a reliable way, at least right now. And there is no incentive for online services to try to figure it out. So there is, in addition to our concerns for lawful fake content, there are also concerns for real journalistic content that the people depicted just don't like. Yeah, I mean, I could see this with like a politician or I did an interview with this woman, lives of TikTok last year, which her team did not like. And, you know, I got a copyright strike, but I was able to appeal that right away and restore it. Of course, that was not generative AI, but you could see a world in which the person who's featured in a video like this would just flag the video as AI of their like this.

And as you mentioned, it would just be taken down. There's no punishment, right, for the platforms if they take down actual real content, is there? There's no punishment for the platforms that they take down real content. I think there are, if I recall correctly in the bill, there are potential penalties for knowingly false notices. but it really seems unlikely that we're going to ever get there, given the way the rest of the bill is structured so heavily towards takedown and so much against users' ability to try to get legitimate content put back up. Right. Well, journalists and other people don't have hundreds of thousands of dollars to even, like, litigate that in court. Exactly. And by the time the video is restored,

it's likely not newsworthy. Who are the people behind this act? And what do you think their motivations are? The best question to ask here is, I mean, this bill seems so bad. So who benefits from it, right? As I mentioned earlier, I think, like, very big holders of intellectual property rights and licenses to people's voices and likenesses are probably going to benefit. So in other words, record labels and movie studios and other big entertainment industry players, which probably explains why they all very strongly supported. And if you can imagine going forward, if they don't already have these types of provisions in contracts, there are going to be provisions and contracts saying, we have license to your face and voice for however long the bill permits them to have it.

And then they can, now that they own the license to that content, they can create as much of it as they want outside of the scope of no fakes and outside of the permission of the actual person being depicted and they can deploy their team of lawyers to create so many automatic takedown requests that anything outside of what they make money off of will be removed from the internet. So that's that's one group. And then another group I can think of, you know, major recording artists would also probably have a good incentive. So if we remember a couple of years ago, there was an AI generated song distributed online and it was called The Heart on My Sleeve. It was made to sound like Drake and the Weekend made it.

And it immediately went viral. It was very close to actually entering the music charts before all of the digital service providers started to take it down. And we assume, based on certain notifications that we saw, that it was coming down on the basis of copyright infringement allegations, probably from their labels. But there's also an incentive for those folks here to have additional protections against this type of what they think is,

infringement on their persona. But the thing about that is, and I don't want to go on too much longer, is there are already rights associated with that. There are already causes of action that could be used. Right. I mean, we didn't have this law two years ago, and it seems like it worked out fine. And there are so many protections. I feel like record labels already have so many protections already over their copyrighted work. It seems gratuitous. Do you think that this will affect content, too, that's not even directly using a person's likeness, but it might be similar? Almost certainly. Again, the scope of the bill technically does not cover, I don't think, like people that don't really exist, but may look similar to real individuals. But the way the

bill is structured, if you think it kind of looks like you, you do have an incentive to file a claim under this because the way the bill is structured, it's almost certain to come down. And that's what you want. That's the result that you want. And users are very unlikely. The users who created whatever has been taken down are very unlikely to have access to the money it would take to file the lawsuit to put it back. So yeah, I do think so. This video is brought to you by Delete Me. As a high-profile journalist dealing with stalking,

death threats, harassment, and more, I know how crucial it is to protect your privacy online. Unfortunately, no matter how careful you are, data brokers will collect and sell your personal information to anyone willing to pay for it. This includes your name, address, phone number, and more. When someone buys your data through a broker, they don't just see you.

They see your spouse, your parents, your loved ones, anyone you're connected to. Removing this information off the internet helps keep you safe from things like identity theft and stalking. But how do you remove your information from the web when there are thousands of data brokers? Delete Me is a subscription service that will automatically remove your personal information being sold online. It's been an absolute game changer for me as somebody that values privacy. Delete Me scours the internet for your information and handles takedown requests. They provide regular privacy reports so you can see how much data they found on you, where it was found, and where it was removed.

You can even make custom removal requests. Their family plans are especially crucial because when you're targeted, your loved ones usually are too. Get 20% off DeleteMe Consumer Plans when you go to JoinDeletMe.com slash Taylor 20 and use promo code Taylor 20 at checkout. That's Join DeleteMe.com slash Taylor 20 and use code Taylor 20 at checkout. Don't wait until it's too late to safeguard your privacy. Make sure your data and your families stays protected. When the content is taken down under this No Fakes Act, does it have to stay down? Or can somebody just re-upload it, for instance, if it gets struck?

The bill creates what we're calling a notice and stay down process, because online services are also required to operate systems to make the content that they have been notified as a digital replica stay down. So that's going to mean content filtering. And that can have some significant problems. So for example, Europe tried to require that kind of content filtering for terrorist materials a few years ago. And even human rights officials at the U.S.

when spoke up about the free expression problems that that would have created. And U.S. groups like mine, like the Center for Democracy and Technology and the ACLU, came out in strong opposition. But the number of pictures and videos threatened by those filters was actually tiny in comparison to what no fakes is going to do. And then the other thing to know about content filters is they're really expensive if you want to do them well. So YouTube in the copyright context has content ID. That cost YouTube something like $100 million to create. Smaller companies don't have that money. So that means they're either going to use clumsy filters or they may find compliance costs too significant to continue at all.

So that's that's another big change in this bill. So with like the filtering, basically it sounds like any content that's even like reminiscent might just get swept up in these filters and taken down because it's reminiscent of some sort of work that's already been, you know, attacked by the no figs. Yeah, it's going to depend on the filtering tool that's used. Some of these tools are direct matching. so if you only get the direct match, those are actually generally kind of pretty accurate. But even so, the problem there is actually going to be context. So you have an unauthorized use of a digital replica, but then over here you have an authorized use. So you have an unauthorized use in a commercial context where you can take, where you can require it be taken down without a problem under the bill.

But over here you have a journalistic one, or you have a use in scholarship and should be should fall under an exception. under the content filtering, even with how accurate the mashing is, they both come down. So that's another big issue here. It seems like just more online censorship of content on across the web. An extraordinary amount of incentivized online censorship, absolutely. That will enable politicians and other people of power to silence dissent and criticism. I want to talk about kind of like the climate that all of this stuff is being proposed under. I talked about the Take a Down Act a little bit.

and I've covered other laws like the Kids Online Safety Act which I hate and other things. But it seems like we're just seeing a lot of legislation proposed that could restrict speech and more recently sort of under the guise of cracking down on AI. Can you talk a little bit about like the legislative climate and how this sort of like boom in it AI has led to more laws kind of like this? Because it seems like we're seeing a lot of this type of stuff crop up. We absolutely are.

And I think this is not a new problem. problem. Impersonation, faked media, non-consensual use of name and likeness, they aren't a new problem. And in the context of COSA, which you brought up, suicidality is not a new problem. Depression and anxiety are not a new problem. But generative AI, I think, triggers this idea of scale even more than social media ever did, right? It is one thing to think of somebody out there spending eight hours on Photoshop trying to create a basically unrealistic non-consensual intimate image of Taylor Swift. Now you have a tool that can create something really realistic within seconds and it can be out there and propagating across the internet seconds later, right? That's what I think we're getting at here. And the fact that that is true, the fact that

this enables harm at a greater scale, it's something we should be worried about. It's absolutely something we should be worried about and thinking about how to address. But we have to do it in the way we've always done it before, which is balanced against the positive uses of this technology and balanced against everybody's human right to free expression. And that is what I see missing in a lot of the legislative proposals that we see, which are very fear-driven and don't do enough to account for constitutional rights. I mean, I think a lot of these people in Congress probably don't use AI very often. And there's just so much like anti-AI consumer sentiment Like in some sense, you have like all the tech people who are totally AI-pilled.

But then, I mean, I hear this a lot. My own reporting of people that are like, just don't use generative AI. And, you know, you won't have to bother with it. And no one should use generative AI anyway because it's so bad and evil and da-da-da. I guess I'm wondering, like, do you think that that like consumer sentiment and the sort of frustrations with AI is also feeding this? Oh, I think absolutely. And in some ways, sort of rightfully so.

If we don't have guardrails around around the use of generative AI in certain contexts, like when it's used to hire somebody, or when it is used to determine somebody's credit rating or what it's used to determine somebody's interest rate, when we don't have guardrails about how it's used or how it's designed in order to ensure fairness in that process, that makes total sense for everybody to have significant desire to do something about that. We should be doing something about that. And we should be doing something about harms with respect to people's reputation, where generative, of AI is involved, but as I said earlier, like, this isn't our first rodeo with this sort of thing.

We have had laws in the past, and we continue to have them. They continue to exist that protect people's right of publicity, that protect people from intentional infliction of emotional distresss. There are ways to address the harms that we see coming, the new ones, and to adapt our ways to address the harms that have already existed to ensure that generative AI is covered appropriately. The way that we're going about it is a little bit different, though. I think the way that we're going about it is to attack the technology itself and try to regulate the technology itself broadly in manners that are going to affect free expression and privacy

far more than I think people intend or want. What is the status of this law? How far has it made it in Congress and kind of what are the next steps for it? It hasn't made it very far yet. It has been introduced, but it has bipartisan support in the Senate. And I believe it has a House companion. Next step would be for there to be a hearing on the bill. And actually, I think there was a hearing on the bill a while ago.

So if there was already a hearing, and I think actually there was, the next step would be for the Senate Judiciary Committee to mark the bill up. At that point, that is a chance to make a bunch of changes to the bill, to try to make it better from the perspective of the legislators after it gets marked up. If it is approved after markup by the Senate judiciary, it goes to the Senate floor, at which point it could pass the Senate. So we still have a ways to go. There is still time to talk about this bill and to talk about the issues with it. And I'm really glad you're doing this as a result. In the CDT release, you guys wrote that this law could affect children and adults who find

themselves separated from power over their own likeness for up to a decade without the appropriate safeguards to prevent abuse. How would that work? The right that this creates, it's licensable. You can give it to someone else. And there are special protections for kids in the bill for licensing, but the odds that, you know, child influencers or their parents truly being able to negotiate to protect themselves in deals with gigantic Hollywood companies just don't seem all that great. And the bill doesn't do nearly enough to protect against that possibility. Well, what should people do that want to kind of express their concerns over the way

that this bill would censor, like journalistic speech, like historical, you know, academic speech, just creative speech generally, like, is there anything that they themselves can do to kind of make a difference? Write to your congressperson. Tell them you're concerned. Call them. Leave them a message. Let them know that the No Fakes Act is not the way to address digital forgeries or digital

replicas. One thing that Daphne says is when she's been taking a look at it is that the no-fakes acts is essentially capitalizing on the reasonable fears that people have related to emerging, emerging in order to ram through over broad restrictions that will harm free expression online. That's exactly right. That's exactly what this is. We were just talking about reasonable fears that people have. And what this bill is doing is essentially using that as an opportunity to give the record industry and movie studios a very big win. And it will come at the expense of journalists, creators, comedians, parents.

and just general online service users. Well, Kate, thank you so much for joining me today. Thank you, Taylor. It's been a pleasure. Thanks again to delete me for sponsoring this video. To get your data removed from the internet, check out join deleteme.com and use code Taylor 20 at checkout for 20% off. All right, that's it for this week's Free Speech Friday. Don't forget to subscribe to my tech and online culture newsletter,

usermag.co. That's Usermag.co, where I write about all of these issues and more. You can watch full episodes of Free Speech Friday every week on my YouTube channel And if you like the show, don't forget to give us a rating and review on Apple Podcasts, Spotify, or wherever you listen. I'll be back next week with a brand new episode of Free Speech Friday. See you then.