Pod Insights Decoder AI deepfakes are cheap, easy, and coming for the 2024 election About this episode Our new Thursday episodes of Decoder are all about deep dives into big topics in the news, and this week we’re continuing our mini-series on one of the biggest topics of all: generative AI. Last week, we took a look at the wave of copyright lawsuits that might eventually grind this whole industry to a halt. Those are basically a coin flip — and the outcomes are off in the distance, as those cases wind their way through the legal system.
A bigger problem right now is that AI systems are really good at making just believable enough fake images and audio — and with tools like OpenAI’s new Sora, maybe video soon, too. And of course, it’s once again a presidential election year here in the US. So today, Verge policy editor Adi Robertson joins the show to discuss how AI might supercharge disinformation and lies in an election that’s already as contentious as any in our lifetimes — and what might be done about it.
Links:
How the Mueller report indicts social networks
Twitter permanently bans Trump
Meta allows Trump back on Facebook and Instagram
No Fakes Act wants to protect actors and singers from unauthorized AI replicas
White House calls for legislation to stop Taylor Swift AI fakes
Watermarks aren’t the silver bullet for AI misinformation
AI Drake just set an impossible legal trap for Google
Barack Obama on AI, free speech, and the future of the internet
Credits:
Decoder is a production of The Verge and part of the Vox Media Podcast Network.
Today’s episode was produced by Kate Cox and Nick Statt and was edited by Callie Wright.
The Decoder music is by Breakmaster Cylinder.
Learn more about your ad choices. Visit podcastchoices.com/adchoices
Transcript 0:01 Speaker 00 Company spending is complex. Managing it should be simple. This is AMEX Corporate, a single platform connecting cards, expenses, and payments to streamline the way you work. Flexible controls for when priorities shift, support around the clock, all backed by American Express, so you can stay focused on what matters most. Built to simplify, designed to scale. This is AMEX Corporate. Get started at go.amex/ammexcorporate. Terms apply.
0:36 Speaker 01 Hello and welcome to Decoder. I'm Neelai Patel, editor in chief of The Verge, and Decoder is my show about big ideas and other problems. Our new Thursday episodes are all about big topics in the news. And this week, we're continuing our short series on one of the biggest topics of all: generative AI. Last week, we took a look at the wave of copyright lawsuits that might eventually grind the entire generative AI industry to a halt. But those lawsuits are basically a coin flip, and the outcome is often a distance as those cases wind their way through the legal system. A bigger problem for the industry right now is that AI systems are really good at making just believable enough fake images and audio. And on top of that, it's an election year here in the United States.
1:20 Speaker 01 The conversation around media manipulation on social platforms comes and goes with every election cycle lately. The 2016 election was marked by Russian disinformation campaigns on Facebook. The 2020 campaign offered a reckoning over the Hunter Biden laptop story on Twitter, and then the major platforms all banning Donald Trump. Those bursts of attention have receded with little or nothing to show for it in terms of regulation. And in the case of Twitter, there's been a wholesale retreat from any moderation at all as Elon Musk has turned the platform into what's now X. And X is where fake pornographic images of Taylor Swift have been most widely distributed. A preview of the AI-related problems now facing every major platform.
2:05 Speaker 01 Dirty political tactics powered by AI are already in the mix. There was a fake Joe Biden robocall in New Hampshire just last month. There have always been lies in campaigns, but that fake Biden robocall was made by a magician from New Orleans who claims it took him just 20 minutes of time and a dollar to generate the fake audio, which is just a glimpse at how AI makes it easy to lie more and lie faster. So today, Verge Policy Editor Addie Robertson joins the show to discuss how AI might supercharge disinformation in an election that's already as contentious as any in our lifetimes, and what exactly might be done about it.
2:52 Speaker 01 Addie Robertson, how are you doing?
2:53 Speaker 02 Hi, good.
2:53 Speaker 01 You've been tracking this conversation for a very long time. It does seem like there's more nuance in the disinformation conversation than before. It's not just Russia made people elect Trump, which is, I think, where we were in 2016. Can you just give a background? What's the shape of how people are thinking of disinformation right now?
3:11 Speaker 02 We've had, I think, about three major in the US presidential election cycles where disinformation was a huge issue. 2016, where there was a lot of discussion in the aftermath about, all right, was there foreign meddling in the election? Were people being influenced by these coordinated campaigns? There was 2020 where deep fakes technically did exist, but generative AI tools were just not as sophisticated. They were not as easy to use. They were not nearly as prevalent. And so there was a huge conversation about what role do social platforms play in preventing general manipulated information? And there was,
3:51 Speaker 02 in a lot of ways, a huge crackdown. There was the entire issue of Stop the Steal. There are these large movements that are trying to just lie about who won the election. What do we do? There were questions about, all right, do we kick Trump off social networks? These were the locus of debate. And now it's 2024, and we have, in some ways, I think, a little bit of a hangover from 2020, where platforms are really tired of policing this. And so they're dealing with, all right, how do we renegotiate this for the 2024 election? And then you have this whole other layer of generative AI imagery, whether or not you want to technically call it deep fakes is like an open question.
4:34 Speaker 02 And then there are all the layers of how that gets disseminated and whether that turbocharges a bunch of issues that already existed.
4:44 Speaker 01 So the platforms are getting tired of this is worth talking about for one second longer. There was a huge rush of how do we make ultra sophisticated content moderation systems? And I think the pinnacle of that rush was Facebook setting up its Oversight Board, which is effectively a Supreme Court of content moderation decisions. And that was seen as, okay, Facebook is as big as a state. It has the revenue of a state. It's a government now. It's going to have some government like functions to regulate speech on its platforms. That didn't pan out, right? Like the Oversight Board exists, it moves very slowly. It's, I think, hard for the average Facebook user or average Instagram user to think
5:26 Speaker 01 there's a moderating force on content moderation on this platform. It's the same as it ever was for the user perspective, as far as I can tell.
5:34 Speaker 02 Yeah, I think that the Oversight Board, what it tends to do is that is maybe comparable to the Supreme Court is do sophisticated outside thinking about what does a consistent moderation framework look like? But like the Supreme Court in real life does not adjudicate every single complaint that you have. You have a whole bunch of other courts. Facebook doesn't have really those other courts. Facebook has a gigantic army of moderators who don't always necessarily even see its policies. So yeah, it's this very macro level, we're going to do the big thinking. But also even at the time, there was the question of, is this really just Facebook or now Meta kind of outsourcing and kicking the can out of its court and putting the hard
6:19 Speaker 02 questions on other people?
6:21 Speaker 01 I wanted to bring that up specifically because that was the pinnacle, I think, of the big thinks about content moderation. Since that time, the companies have all done lots of layoffs. We've seen trust and safety diminished across the board. I think most famously with Twitter, now X, Elon Musk basically decimated the trust and safety team on that platform. It appears Lindy Acarino is trying to bring some of them back. But the idea that content moderation is the thing these platforms have to do has, is no longer in vogue, I think, the way it was when the Oversight Board was created.
6:56 Speaker 02 Yeah, and part of this is also political that there was a huge largely in the again in the US right wing backlash to this, that this was the kind of thing that would get a state attorney general mad at you and get a congressional committee to investigate you as it ended up doing with pre Musk Twitter. I think that yeah, there became a real political price for doing this as well. Since then, some platforms have let Donald Trump back on. They've said, all right, but we cannot possibly moderate every single lie on this. We're going to just wash our hands of whether you're saying the election was stolen or not.
7:33 Speaker 01 So yeah, let's go through the new players and how they might turbocharge the disinformation conversation now. And then let's talk about what might be done about it. I do just want to emphasize for the audience, it doesn't seem like the desire to regulate information on social networks is nearly as high as it has been in the past. And I think that is an important thing to start with because the technical challenges are so hard that wanting to solve them is actually an important component of the puzzle. But let's talk about the actual technical challenges and the players behind them. OpenAI, that's a new company. There are a lot of other new companies in various stages of controversy. So Midjourney
8:15 Speaker 01 exists, that is an image generator. Stability AI exists, another image generator. They're getting sued by Getty for using the Getty library allegedly to train images that look like Getty photos. In this context, very important. Midjourney is getting sued as well. OpenAI is getting sued for training on the New York Times database. Just a few days ago, OpenAI announced Sora, its text to video generator, which frankly makes terrifying videos. All those videos are terrifying. But you can see how An enterprising scammer could immediately use that to make something that looks like compelling video of something that didn't happen. All of these companies talk about AI alignment, making sure AI doesn't go off the rails. Where's the AI industry broadly on we shouldn't do
9:00 Speaker 01 political deep fakes? Do they have a unified point of view? Are they all in different spots? How's that working out?
9:05 Speaker 02 The companies are in slightly different spots, but they actually have come together. Very recently, they've signed an accord that says, look, we're going to take this seriously. They've announced policies that are varying levels of strictness, but tend toward if you're a major AI company, you're going to try to prevent people from creating information that maybe looks bad for public figures. Maybe you ban producing images of recognizable figures altogether, or you try to. And you have something in your terms of service that says if you're using this for political causes, or if you're creating deceptive content, then we can kick you off.
9:44 Speaker 01 One challenge here in America is the existence of the First Amendment. The Biden administration recently did an executive order saying, don't do bad stuff. And these companies all agreed, okay, we won't do bad stuff. But the United States government is pretty restricted in saying you can't make deep fakes of other people because the First Amendment exists and it can't control that speech directly. Are the companies rising to that challenge? It's like we will self-regulate because the government can't directly regulate us?
10:13 Speaker 02 We don't know necessarily how good the enforcement of it is going to be, but the companies seem so far pretty open to the idea of self-regulation, in part because I think this isn't just a civic-minded political thing. Dealing with unflattering stuff about real people is just a minefield they don't want. That said, there are also just there are open source tools. Stability AI is pretty close to open source. It's pretty easy to go in and make a thing that builds on it that maybe strips away the safeguards you'd get in its public version. So it's just not quite equivalent to the way that, say, social platforms are able to completely control what's on their platforms.
10:57 Speaker 01 So you've got a handful of companies with varying sets of restrictions, a broad general industry consensus, we shouldn't do deep fakes. And then you have reality, which is that there are deep fakes of celebrities all the time. There are deep fakes of teenage girls in high schools that are getting circulated on private message boards. It is happening. What can be done to stop it?
11:18 Speaker 02 Does stopping mean that you're just trying to limit the spread to where this doesn't become a huge viral thing that a bunch of people see, but it's still maybe technically possible to create this? Or do you want to say, all right, we have a zero tolerance policy if anything is created with any tool anywhere, even if someone keeps it to themselves that is unconscionable?
11:38 Speaker 01 Let's start with the second one, which I think has the more obvious answer. Saying no deep fakes are allowed whatsoever seems like it comes with a host of unintended consequences about speech, and also seems like impossible to actually accomplish because of the existence of open source tools. Like, I think, how would you actually enforce a total ban on deep fakes? And the answer is that Intel and Apple and Qualcomm and Nvidia and AMD and every other chipmaker have to prevent it somehow at the hardware level, which seems impossible. The only example I can think of where we have allowed that to happen is that
12:19 Speaker 01 Adobe Photoshop won't allow you to scan and print a dollar bill, which makes sense. Like, it broadly makes sense that Adobe made that deal with the government. But it's also like, well, that's about as far as you should let that go, right? Like, there's a point where you want to make a parody image of a Biden or a Trump, and you don't want Photoshop saying, hey, are you manipulating a real person's face? Like you're saying, that seems way too far. So a total ban seems implausible. There are other things you could do at the creation step. OpenAI bans certain prompts that are violates their terms of service. Getty won't let you talk about celebrities at all. If you type a celebrity's name or basically any proper noun into the Getty image generator, it just tells you to go
13:00 Speaker 01 away. There's a lot of conversation about watermarking this stuff and making sure that real images have a watermark that say they're real images and AI images have a watermark that say they're AI images. Do any of those seem promising?
13:14 Speaker 02 The most promising argument I've heard for these is the idea that you can, and this is an argument that Adobe has made to me, train people to expect a watermark. And so if what you're saying is we want to make it impossible to make these images without a watermark, I think that raises the same problems that we just talked about, which if anyone can make a tweaked version of an open source tool, they can just say, don't put a watermark in. But I think that you could potentially get into a situation where you require a watermark. And if something doesn't have a watermark, there are ways that its design, or its spread, or people trusting it are severely hobbled. That's maybe the best argument for it I've heard.
13:58 Speaker 01 The part where you restrict the prompts, OpenAI restricts the prompts, Getty restricts the prompts. It's pretty easy to get around that, right? The Taylor Swift deepfakes that were floating around on Twitter, they were made in a Microsoft tool, and Microsoft just had to get rid of the prompts. Is that just a forever cat and mouse game on the restrict the prompts idea?
14:16 Speaker 02 It does seem like the thing about a lot of generative AI tools is that there are just vast, vast numbers of ways to get them to do something. People are going to find those. Software bugs are a thing that has been a problem. Zero day exploits have been a problem on computers for a very long time. And this feels like it kind of falls into that category.
14:39 Speaker 01 That's the creation side. We need to take a quick break. When we come back, we'll get into the harder problem, distribution.
14:49 Speaker 03 Are your ad campaigns lighting up the dashboard, but not the pipeline? That's bullspend, and marketers are calling it out in. Dashboard, confessions.
14:59 Speaker 04 My boss asked for results, so I opened my dashboard for the only positive sounding metric I had: impressions.
15:06 Speaker 03 Cut the bullspend. See revenue, not just reach. LinkedIn delivers the highest return on ad spend of major ad networks. Advertise on LinkedIn. Spend $250 on your first campaign and get a $250 credit. Go to LinkedIn.com/campaign. Terms and conditions apply.
15:23 Speaker 01 Welcome back. So we've talked about what the companies that make software and hardware can do about the creation of deepfakes. And it seems like the best answer we have right now is adding watermarks to AI-generated content. But the real problems are in distribution.
15:38 Speaker 01 Let's talk about the distribution side, which is, I think, where the real problem lies. If you make a bunch of deepfakes at your house with Donald Trump, and you never share them with anyone, what harm have you caused? You start telling lies about both presidential candidates, and you share them widely on social platforms, they go viral. Now you have caused a giant external problem. And so it feels like the pressure to regulate this stuff is going to come to the platforms. And again, I think the desire of the platforms to moderate waxes and wanes, and it feels low right now, maybe it'll ramp back up. Where are the platforms right now with the deep fake distribution problem?
16:15 Speaker 02 So far, it feels like the consensus is, we're going to label this, and that's going to be mainly our job, is that we're going to try to make sure we catch it. There are cases where, say, maybe you get it taken down if you haven't disclosed, if you're a company, or you're buying a political ad. But broadly, the idea seems to be, we want to give people information and tell them that this is manipulated, and then they can make their own call.
16:42 Speaker 01 The one platform that stands out to me, and you and I have talked about this a lot, is YouTube, which has an enormous dependency on the music industry. The music industry is not happy about AI-generated covers using the voices of its artists. Notably, fake Drake caused a huge kerfuffle. Universal Music Group went to Google, they announced some rules, they're going to prevent deepfakes, or allow some licensing of the money to flow back to the artists. That is a very private sort of licensing scheme that sits outside of the law, it sits outside of the other platforms. Do you think YouTube is going to lead the way here because it has that pressure? Or is that just a one-off to the music industry?
17:23 Speaker 02 I feel like the incentives for something like the music industry, and for things that are basically aesthetic deepfakes, I think the incentives there are very different than they are for political manipulated imagery. That a lot of the question with YouTube is, okay, you are basically parodying someone in a way that may or may not legally be considered parody. And we can make a deal where that person really, all they want is to get paid. And maybe they want something sufficiently controversial taken down. But if you give them some money, they'll be happy. That's just not really the issue at hand with political generated images. The problem there is around reputation.
18:03 Speaker 02 It's around people who do, at least in theory, care about, did this person say this thing? Is this true? So I just, I don't know that you could cut a deal with Joe Biden that says every time Joe Biden, you make something up about him, he gets a penny.
18:20 Speaker 01 I feel like politicians are always asking for donations. Maybe that's just the way to solve the problem. You just pay for the lies. As long as, as long as politicians are getting paid, from what I gather, I think particularly Donald Trump, as long as he's getting paid, uh, he might be, he might be cool with it. Outside of YouTube, which does have this big dependency on the labels and licensing. And so I think is leading the way on having any particular policy with specificity. Do any of the other platforms have ideas here that are more than we have an existing policy? And we'll see how it works with the problem.
18:50 Speaker 02 There are companies that are signing on to an initiative called C2PA, which is, we were talking about watermarks earlier. It's a content provenance system. It includes a watermark that has metadata. And the goal there is the idea that you will be able to at least tell where something has come from and whether it's manipulated and that it's supposed to be this broad industry wide. Everybody has the same watermark system. So it's very easy to look at an image and pop it in and check and see if it has the watermark. That's one of the leading ways the AI industry at this point is trying to deal with truth and provenance.
19:29 Speaker 01 Is that shipped yet? I feel like we've been talking about C2PA and the Content Authenticity Initiative. We've had Dana Rao, Adobe's General Counsel on the show. He said the deep fake of the Pope wearing a puffer jacket was, quote, "a catalyzing event for content provenance", which is an amazing quote and all credit to Dana for it. But there's people wanting to do it. There's the activity we see and then there's shipping it. Has that shipped anywhere? Can you go look at it?
19:53 Speaker 02 Watermarks are rolling out places. OpenAI adopted them in mid-February. They're starting to appear on Dolly images. You can look at them in Photoshop. I think the problem is more that this thing rolled out, but really most people are not going to care enough to check.
20:10 Speaker 01 Well, unless the labels are in their face, right? Unless you are scrolling on TikTok and you see something and TikTok puts a big label right over the top of it that says this is AI, which doesn't seem to be happening anywhere. You know, OpenAI did Sora, its video generator. The videos are compelling, although they have some extremely terrifying errors in them. But there's not like a big label on them that says this is AI generated. Like, they're gonna travel without the context of OpenAI having produced them to promote their AI tool. And even that seems dangerous to me.
20:43 Speaker 02 Yeah, a lot of the issue with C2PA right now is that you have to actually go in and pop it into a tool to check the metadata, which is just an extra step that the vast majority of people are not going to take. And that, yes, it's not applying to things like Sora yet, at least as far as OpenAI has told us. So there is not a really prominent in-your-face, this thing is AI in most cases.
21:09 Speaker 01 Can you remove those watermarks?
21:11 Speaker 02 I mean, a screenshot tool, as far as I can tell, can remove the watermarks. And I think there are ways that you can end up just stripping these things out. It's very, very hard to create a perfect watermarking system. Just so much of this relies on a lot of Adobe's argument being, well, eventually, we want people to expect this, that it's gonna be like, you know, if you look in your browser and you get a certificate warning that says there's no certificate for this webpage, maybe you won't trust the webpage. I think the goal they're going for is the idea that everyone will be trained into expecting a certain level of authenticity, and I just don't think we're at that point. In some ways, these problems already existed. Photoshopped nudes have been a thing that has been used
21:54 Speaker 02 to harass people for a very long time. And photoshopped images of politicians and manipulated content about politicians is nothing new. A thing AI definitely does is scale up those problems a huge amount and make us confront them in a way that it was maybe easier to ignore before, especially by adding a technology that...
22:15 Speaker 02 The people who are creating the technology are trying to hype up in a way that sounds terrifying and world ending for other reasons. The problem with a lot of this is that you can't apply the kinds of paradigms that social media has because it really only takes one person with one capability to do a thing. Like, it takes one bad actor to make something that you can spread in huge variations that are hard to recognize across huge numbers of platforms. I think that raises slightly different problems than, say, there was this big account on social media that's spreading something. Well, all right, Facebook can ban them.
22:57 Speaker 01 So, we've talked a lot about what the platforms can do, what the AI companies can do, as private companies, these initiatives like content authenticity, that's a private framework. The government has some role to play here, right? The big challenges are the First Amendment exists in the United States, and that really directly restricts the government from making speech regulations. And then, you know, we are a patchwork of state and federal laws. What is the current state of deep fake law? Are you allowed to do it? Are you not allowed to do it? How's that work?
23:30 Speaker 02 It's mostly a small patchwork of past laws, a huge number of laws of varying likely constitutionality that people are debating, and not a whole lot at the federal level. There are a lot of different problems that AI-generated images pose, and there are cases where individual states have passed rules for individual problems. There are a few states that incorporate, say, non-consensual AI pornography laws into anti-general revenge, non-consensual porn rules. There are a few states with rules about how you have to disclose manipulated images for elections. And there are some attempts in Congress to create a larger framework, or in the, say,
24:13 Speaker 02 FEC and other government regulatory agencies, to create a larger framework. But we just are still in this large, chaotic period of people debating things.
24:24 Speaker 01 Let's start with non-consensual deep fake pornography, which I think everybody agrees is a bad thing, that we should find ways to regulate away. A solution to revenge porn broadly on the internet is copyright law. Right? You have made these files with your phone or your computer. Someone else distributes them. You say, no, those are, those are mine. Copyright law will let me take this down. When you have deep fakes, there is no original. It's not a copy of something that you've made or that you own. You have to come up with some other way to do it. Right? You have to come up with some other mechanism, whether that's just a law that says this is not right, or it's some other idea like the right to your likeness. Where have most of the existing laws landed there?
25:08 Speaker 02 The copyright issue is actually something that came up with non-synthetic, non-consensual pornography because say if one of your partners took a nude picture of you, you don't own that picture. And that was already just a huge loophole that legislators have spent about a decade trying to make laws that meaningfully address non-consensual pornography that's not AI-generated. And the frameworks they've come up with are getting ported over to AI-generated imagery that a lot of it is about, all right, this is harassment, this is obscenity, this is some other kind of speech restriction that is allowable. A lot of non-consensual pornography is a kind of sexual harassment that we can find
25:51 Speaker 02 ways to wall outside protected speech and that we can target it in a way where it's not going to necessarily take down huge amounts of other speech, the way that say just banning all AI-generated images would.
26:05 Speaker 01 There are a bunch of state laws around non-consensual AI-generated pornography. What states are those and is there any federal law on the horizon?
26:14 Speaker 02 There's California, New York is another, there's Texas. At the federal level, there have been attempts to work this into, it's not a criminal statute, but there is a federal civil right to sue if you have non-synthetic, non-consensual porn of view, and there have been attempts to work AI into that and say, all right, well, it's not a crime, but it's a thing that you can sue for under, I believe it is the reauthorization of the Violence Against Women Act. And then there have been attempts to, like you mentioned, just tie all of this into a big federal likeness law. So likeness laws are a mostly state level thing that says, all right, you can't take
26:57 Speaker 02 Taylor Swift and make her look like she's advertised your instant pot.
27:04 Speaker 02 And so there have been some attempts to make a federal version of that, but likeness laws are really tricky because they're so much broader that they end up catching things like parody and satire and commentary. And they're just, I think, much riskier than trying to create really targeted, specific use laws.
27:24 Speaker 01 The idea that someone should be in absolute control of a photograph of themselves has only gained prominence over time. Emily Radikowski wrote that great essay for The Cut several years ago, where she said, "A street photographer took a photo of me and I put it on my Instagram and I'm suing him to say that I can take his photo because it's a photo of me." And that is a very complicated argument in that case. But the idea that you should be in total control of any photo of you, I think a lot of people just instinctively believe that. And I think likeness law is what makes that have legal force, but you're saying, oh, there's some stuff here you wouldn't want to pull under that umbrella.
28:07 Speaker 02 If you're talking about non-synthetic stuff, then there are all kinds of documentaries and news reports and really things that people have a public interest in making where you don't want to give someone the right to say, "You cannot depict me in a thing." And in that case, it's doing something I actually did, but AI-generated images raise the whole other question, which is, okay, so where do you draw the line between an AI-generated image and a Photoshop of someone and a drawing of someone? Should you not be able to depict any person in a situation that they don't want to be depicted in, even if that situation is something that would just broadly be protected by the First Amendment?
28:47 Speaker 02 Like, where do we think that the societal benefit of preventing a particular usage that hurts someone should be able to override the interest we have in just being able to write about or create images of someone?
29:05 Speaker 01 We have to take another quick break. We'll be right back.
29:15 Speaker 03 Are your ad campaigns lighting up the dashboard? But not the pipeline? That's bullspend. And marketers are calling it out in. Dashboard Confessions.
29:25 Speaker 04 My boss asked for results, so I opened my dashboard for the only positive sounding metric I had: impressions.
29:32 Speaker 03 Cut the bullspend. See revenue, not just reach. LinkedIn delivers the highest return on ad spend of major ad networks. Advertise on LinkedIn. Spend $250 on your first campaign and get a $250 credit. Go to LinkedIn.com/campaign. Terms and conditions apply.
29:49 Speaker 01 We're back, talking with Verge Policy Editor Addie Robertson about why it's really hard to limit either the creation or sharing of deep fakes.
29:57 Speaker 01 So that's the philosophical policy debate. You want to restrict this because in many cases, it can be used to do very bad things. There are some things that we absolutely want to forbid, but if we let that get too wide, we're going to start running into people's everyday speech. We're going to start running into absolutely constitutionally protected speech, like documentaries, like news reporting. That's pretty blurry. And I think the audience here, you should sit with that because that is pretty blurry. On the flip side, there are two bills in Congress right now that purport to restrict on this stuff. There's something called the No Fakes Act, which is Chris Coons, Marshall Blackburn, Amy Klobuchar, Tom Tillis. And then after the Taylor Swift situation on X, there's something called the Defiance
30:41 Speaker 01 Act, which stands for the Disrupt Explicit Forged Images and Nonconsensual Edits Act, which is quite a lot of words. Do they go towards solving the problem? Do you see differences there? Do you see them as being an effective approach?
30:55 Speaker 02 The two bills are a little bit the thing I talked about where one of them, the Defiance Act, is really specifically about we want to look at non-consensual pornographic images. We define what that means, and we think that this particular thing we can carve out. There are lots of questions about, in general, how far you want to go in banning synthetic images, but it's really targeting sexually explicit pictures of real people. And I think things like the No Fakes Act, I believe there's also something called the No AI Fraud Act. These are much broader. We just think that you shouldn't be able to fake images of people. And we're going to make some carve outs there, but the fundamental idea is that we want
31:38 Speaker 02 to create a giant federal likeness law. And I think that's much riskier because that is much more a, we start from a point of saying that you shouldn't be able to fake an image of someone without their permission, and then we're going to create some opt-ins, some options where you're allowed to do it. And I think that raises so many more of these questions that do we really want to create a federal ban on being able to create a fictionalized image of somebody?
32:13 Speaker 01 That is the likeness law approach to it, which has big problems of its own. Another approach we've heard about on Decoder is rooted in defamation law. So Barack Obama was on Decoder. He said, "There are different rules for public figures than 13-year-old girls. We're going to treat them differently. We should have different rules for what you can do with a public figure than teenagers. We should have different rules for what is clearly political commentary and satire versus cyberbullying." And then Senator Brian Schatz was recently on, and he said, he said something similar. Is defamation where this goes? Where it's, hey, you made a deep fake of me. Maybe it's, maybe it's my likeness, but you're actually defaming my character, and you did it on purpose.
32:54 Speaker 01 And that rises to the level of you knowingly telling a lie about me, and I can, defamation law is what's going to punish you for this, instead of some law about my likeness.
33:04 Speaker 02 Defamation law has already come up with text-based generative AI, where if something like ChatGPT tells a lie about you, are you allowed to say they're making things up about me? I can sue. And I think the benefit of defamation law is that there is a really huge framework for hammering out when exactly something is an acceptable lie and when it's not. That, all right, well, would a reasonable person believe that this thing is actually true? Or is this really obviously political commentary and hyperbole? I think that we're on at least more solid ground there than we are with just saying, all right, fine, you know what? Just ban deep fakes. I do think that still defamation law is complicated.
33:47 Speaker 02 And every time you open up defamation law, as Donald Trump has once suggested, you end up getting a situation where in a lot of cases, it's very powerful people throwing the law against people who don't necessarily have the money to defend themselves. And in general, I'm cagey about trying to open up defamation law. But it is a place where at least you have a framework that people have spent a very long time talking about.
34:18 Speaker 01 One thing we constantly say here at The Verge is that copyright law is the only real law on the internet, because it's the only speech regulation that everyone just kind of accepts. Defamation law is not a speech regulation that everyone just accepts, that has boundaries, the cases go back and forth. The idea that there should be a federal right to likeness doesn't even exist yet. So that feels like it will be very controversial if it happens as a speech regulation. But at the heart of that is the First Amendment, right? People have such a strong belief in the First Amendment, that saying the government should make a speech regulation, even if something is really bad, is extraordinarily complicated and high barrier to cross. Do you see that changing in the context of AI?
34:59 Speaker 02 When a new technology comes along, there are a large number of people who don't necessarily think about it in terms of the First Amendment or of speech protections, where you're able to say, oh, well, this thing is just categorically different. We've never had technology like this before. The First Amendment shouldn't apply. I always hope we don't go there with the technology because I think that the problems that come from just blanket outlawing it tend to be really huge. I don't know. I think that we're still waiting to see how disruptive AI tech actually is. We're still waiting to see whether it is meaningfully different from something like Photoshop,
35:42 Speaker 02 even though it seems intuitively like it absolutely should be. But we're still waiting to see that play out.
35:49 Speaker 01 We spent a lot of time talking about the visual side of it. We're going to make deep fake images. Those images have real world harms, especially to young people, especially young women. In an election cycle, making it seem like Trump or Biden fell down the stairs could be very damaging. There's also the voice side of it, right? Where having Joe Biden do AI-generated robocalls is a real problem.
36:13 Speaker 01 Or convincing people on TikTok that Trump said something he didn't say is a real problem. Do any of these laws address that aspect of it?
36:22 Speaker 02 If we're talking about non-internet systems like robocalls, then we actually have laws that aren't really even related to most of the things we've talked about. There is a rule called the TCPA that's an anti-robocall law, basically, that says you cannot just bombard people with synthetic phone calls. And it was recently decided that, all right, should artificial voices there include voice cloning? Yes, obviously. So at this point, things like robocall laws apply to AI. And so if you're going to try to get Joe Biden calling a bunch of people and telling them not to vote, that's something that just can be regulated under a very longstanding law.
37:01 Speaker 01 What about a fake Joe Biden, Joe Rogan podcast on TikTok?
37:06 Speaker 02 That raises really all the same questions that image-based AI raises. In some ways, it's probably going to be harder to detect and regulate against at a non-legal platform level because so much stuff is optimized for detecting images. And so in some ways, it's maybe even a thornier problem. And also, on the other hand, voice impersonation was a thing before this, that there were really good impersonators of celebrity voices. And so I think that that might be a technically harder problem to fix, but I think that the legal questions it raises are very similar.
37:45 Speaker 01 All right, so we've arrived at what I would describe as existential crisis. Many, many problems, one set of things that should clearly be illegal, which deep fake, non-consensual pornography seems like it should clearly be illegal. Everything else seems kind of up for grabs. How should people be thinking about these challenges as they go into this election year?
38:04 Speaker 02 There are a bunch of really hard, technical issues. And a lot of those issues are going to be irrelevant to people because so many people do not check even very obviously fake information because of a variety of reasons that do not have anything to do with it being undetectable as a fake. I think that trying to actually make yourself care about whether something is true is in a lot of ways, a bigger, more important step than making sure that nothing false is capable of being produced. I think that's the place where huge numbers of people have fallen down and where huge numbers of people have fallen for things.
38:48 Speaker 02 And I think that while all of these other issues we've been talking about are incredibly important, this is just a big, individual psychological thing that people can do on their own that does not come naturally to a lot of us.
39:05 Speaker 01 Thanks again to Verge Policy Editor Addie Robertson for joining us on Decoder. These issues are so challenging, and she always helps me understand them so much more clearly. From now on, we're going to keep bringing you second episodes of the show every Thursday to deliver more analysis and storytelling like this, in addition to our classic, regular, weekly interviews with CEOs, lawmakers, and other troublemakers. Stay tuned for part three of our AI series next week. If you have thoughts about this episode or what you'd like to hear more of, you can email us at [email protected] . We really do read every email, and we talk about them quite a bit. You can also hit me up directly on Threads on thatReckless1280. We also have a TikTok, it's a lot of fun, check it out, it's at DecoderPod. If you like Decoder, please share it with your friends, subscribe wherever you get your podcasts, if you really love the show, hit us with that five-star review. By the way, we really want reviews for these new second episodes. Go do it.
39:50 Speaker 01 Decoder is a production of The Verge and part of the Vox Media Podcast Network. Today's episode was produced by Kate Cox and Nick Statt, it was edited by Kelly Wright. The Decoder music is by Breakmaster Cylinder. We'll see you next time.
40:07 Speaker 05 Everyone's got a hard question about AI. Your job, your kids, where it's going. Anthropic was built to surface those questions and share what it finds along the way, because there's hope in hard questions. Ask yours at claud.ai/spotify and keep thinking. That's claud.ai/spotify.
Previous Crunchyroll President Rahul Purini on how anime took over the world Next Guest host Hank Green makes Nilay Patel explain why websites have a future Page updated October 26, 2026 · transcript and summary are machine-generated