A (3:19)

Let's do it. This is what Roman Mars can learn about Con Law. An ongoing series of indeterminate length and now somewhat predictable release where we look at deep fakes memes and the lying liars who spread them and use them to examine our Constitution like we never have before. Our music is from Doomtree Records, our professor and neighbor is Elizabeth Jo, and I'm your fellow student and host, Roman Mars.

B (4:11)

This episode is brought to you by Progressive Commercial Insurance Business Owners Meet Progressive Insurance. They make it easy to get discounts on commercial auto insurance and find coverages to grow with your business Quote in as little as 7 minutes@progressivecommercial.com Progressive Casualty Insurance Company coverage provided and serviced by affiliated and third party insurers. Discounts and coverage selections not available in all states or situations.

A (4:41)

So let's start with what is a deepfake?

B (4:44)

Okay, so of course people have heard of photoshopping images, and the Lincoln portrait tells us that it was possible, even 150 years ago, to make a convincing but intentionally altered image. But when researchers on the media or computer scientists use the word deep fake, they're referring to the use of a specific kind of artificial intelligence, and it's called generative AI. And this generative AI is used to create new pictures, text, audio or video that didn't exist before, but looks very close to the real thing, so much so that you might be fooled into thinking that it is real. And without getting too much into the details here, these programs can create these very realistic images because the AI involved in them has been trained on or learns from huge amounts of information that already exists, like videos, audio, clips, most of it that's available on the Internet. So you have a very sophisticated tool, but you don't have to be an expert to use it. And a lot of people are probably familiar with ChatGPT's large language model, which came out about two years ago. So with ChatGPT, you can just type in an ordinary question or a command and you can use it to write a sentence. In fact, you can write whole paragraphs or even essays without doing anything yourself. You don't need any special knowledge to have the program produce these results. Now, the same is true for creating a deepfake. It's very easy to find tools online to create very realistic audio or video just by providing written commands. Now, deepfakes have already been used in some very disturbing ways. People have inserted real faces, almost always women, without their permission, on sexually explicit photos or videos and then share them online. These manipulated images, they're sometimes called revenge porn. It's a huge problem. There's also the problem of using deepfakes to commit fraud. And some people are worried that we could have a national security emergency or manipulation of the stock market, all kinds of crazy scenarios that could happen. But elections are another way that deepfakes can be very, very harmful.

A (6:56)

So what examples have we seen of that?

B (6:58)

Well, there have already been several election deepfakes that have circulated. Earlier this year, when we saw the clash between Texas and the Biden administration over immigration enforcement, there was an audio posted on sites like TikTok that claimed to be a recording of President Biden. And the voice in this audio, which sounds very much like President Biden says, if we have to send F15s to Texas here and wage war against Texas, so be it. It was a deep fake, not real. And before that, in March of 2023, right before Trump was indicted on New York state criminal charges over his hush money payments, pictures started to appear online. And these pictures show Trump being arrested. Another one shows him being dragged away by the police. And in response, Trump posted a picture of himself kneeling in prayer. All of these pictures were deep fakes. None of them were real, but they did look real. In an election, we worry that deepfakes can be created and shared to damage a candidate's reputation and to deceive voters to influence elections. Again, you don't need any special knowledge to create these videos or audio clips. You can just type in ordinary words about what you want and the programs will generate the result. So that means, Roman, you could type in or prompt an image generator with something like create a photo of Kamala Harris taking illegal drugs or create a photo of voter ballots in Pennsylvania being thrown in a dumpster or a photo of armed militia groups at an Ohio polling place. So you get the idea. Things that aren't real but look very realistic and could change voters minds. Right now most of these deepfakes are pretty quickly identified and sometimes by experts, but sometimes it's obvious to anybody who takes a close look. So I don't know if you've seen any of these pictures, but it's common today for AI generated photos to get faces right, but other parts of their bodies are strange. So there was a widely circulated picture of Trump posing with a group of Black women. And at first, if you look at it quickly, it looks pretty realistic. But if you look closely, some of the people in the background are missing their fingers. It was a deep fake, but that's right now, and the technology is improving all the time. And so with elections, I think the problem is this. What happens when there's a deep fake that just might be real, and maybe it comes out just a day or two before the election? And what if there isn't time for the candidate in that deep fake to deny that the picture is real, to say, this is a fake, I'm not in it? And then what if a viral deep fake does impact the election in ways that we can't do anything about it?

A (9:51)

I mean, so, given that, are deepfakes illegal in any way?

B (9:55)

Well, yes and no, but mostly no. There is a lot of interest by states and members of Congress trying to figure out what to do about election deepfakes. But right now, there's no federal law that bans deepfakes or regulates them when they're used to influence elections. And most states don't regulate election deepfakes, but a small number of them do. There's a bunch of proposed state and federal laws, but they're certainly not going to apply in time for this presidential election.

A (10:25)

So of the places that do have some laws against deepfakes, how do they regulate them?

B (10:30)

So there's a lot of variation in the existing laws that do regulate election deepfakes. There's the issue of what should be regulated. So some of these state laws say that, look, if you're going to distribute a deepfake in a political ad, the image has to state that you're using an altered real image. It has to be transparent that this is a deepfake. But some other state laws go further and ban the use of deepfakes that are used to influence an election. In 2019, California enacted a law that bans anybody from distributing with actual malice a deep fake of a candidate running for office within 60 days of an election. And to violate that law, the person posting the deepfake would have to intend to injure the candidate's reputation. And it has to be the kind of deepfake that the California law says appears to a reasonable person to be authentic.

A (11:26)

That's a lot of caveats.

B (11:28)

A lot of caveats.

A (11:29)

But are those limitations constitutional?

B (11:32)

That is a very good question, especially about bans on election deepfakes, and that's because deepfakes are speech. Yeah, that's right. Even if they are deceptive. Lies. In fact, you and I have talked before about how lies can be protected by the First Amendment. A lie doesn't lose First Amendment protection just because it is a lie. So we can think of election deepfakes as political speech. That's why a law that would ban all deepfakes would probably violate the First Amendment. But that doesn't mean that the government can't regulate political speech. But if it does, any government restriction of a deep fake has to be very, very careful. The government would have to show that a deepfake ban was necessary to prevent a very specific harm, that this ban was really the only way to prevent that harm. So for some of the reasons the government might want to severely restrict election deepfakes would be having an informed electorate. The Supreme Court has said that's a good reason that the government can try to regulate political speech. Maybe that would be a reason for, let's say, requiring that an election deepfake would have to be labeled. The government might also be able to argue that some deepfakes are a form of voter intimidation or undermining the electoral process. But an election deepfake ban, in other words, a law that says no election deepfakes would have to allow some exceptions, too. First, there's the protection of press freedom. Even if we want to stop deepfakes and we think they're harmful, we don't want to stop journalists from reporting on election deepfakes. In 2018, the comedian Jordan Peele, he presented a deepfake video of former President Obama. But the entire point of his video was to raise awareness about deepfakes. It would be weird to ban that particular instance. Right, because it was simply a political deep fake. That wouldn't make sense. Yeah, but, Roman, there's another, maybe less obvious issue with deepfake bans. And that is? Well, aren't some deepfakes just jokes? And couldn't we think of some political deep fakes that are kind of trolling, but they actually have the serious purpose of political criticism? And that's where Hustler magazine comes in. You familiar with that magazine?

A (13:51)

Yeah, yeah, I'm familiar. Yeah.

B (13:54)

Yeah. It's a pornographic magazine.

A (13:57)

And that's also a long and storied legal history as well.

B (14:01)

That's correct. And that's because in the November 1983 issue of Hustler magazine, it featured an ad that was a parody of ads that you'd see for Campari and Italian liqueur. The real ads featured celebrities talking about their first time with Campari. And it was an intentionally suggestive ad campaign. And the Hustler ad portrayed Jerry Falwell talking about his first time. Now, for those who don't know, Falwell was a nationally known televangelist and conservative activist in the 1980s. And he's known for founding the Moral Majority, a political movement associated with the Christian right and the Republican Party. And in the Hustler ad, Falwell describes his first time. The Supreme Court described Falwell's first time as a, quote, drunken, incestuous rendezvous with his mother in an outhouse. Why the supreme court? Because in 1988, the Supreme Court reviewed an appeal brought by Larry Flint, publisher of Hustler magazine. And Flint had been sued by Falwell over this fake ad. Now, Flint argued that his magazine's satirical depiction of Falwell was protected by the First Amendment. And the Supreme Court agreed. Now, Falwell had sued Flint for intentional infliction of emotional distress. But the Supreme Court said Falwell, who is a well known public figure, couldn't recover without showing that the magazine ad had a false statement of fact made with actual malice. And that was the problem for Falwell, right? This was a parody. The Hustler ad was a big joke. Nobody thought that it was really an interview with Falwell. It was just a joke making fun of Falwell's sanctimonious reputation. And so, in the court's view, as long as the magazine couldn't be interpreted as stating actual facts, it was protected First Amendment speech, even if it was offensive. Today we associate that Hustler magazine case with the idea that the First Amendment protects parody and satire. Even though the Supreme Court clearly didn't think much of the Falwell ad in the case itself, it used the case to talk about the importance of satire in political debate. In the case, the court said that the very appeal of political caricatures was that they were usually based on the exploitation of unfortunate physical traits or politically embarrassing events. So you have this case about a lewd parody in a pornographic magazine. But in it, the Supreme Court observed that one of the prerogatives of American citizenship is the right to criticize public men and measures. You said that was a lot of words about the California law. That's why the California election deepfake ban has an actual malice standard. And why the law recognizes exceptions for deepfakes that are intended to be sex, satire or parody.

A (17:03)

I see, I see.

B (17:05)

But that's just California. Some other proposed laws aren't written this way. A recently proposed election deepfake law in Georgia would have made it a crime to share a deepfake with the intent of influencing an election. But the Georgia law had no exceptions for parody or satire. And so presumably that would be a constitutional problem.

A (17:26)

I mean, one of the things that you mentioned when stating the California law, is this, like, 60 days within an election or something having to do with not just the content of the deepfake or the material that's being created, but sort of the time and space that is being used? How do those figures constitutionally, are those a little bit more robust? Do they stand up more, or do those have similar challenges to the idea of intent and deceit?

B (17:52)

Well, I mean, I think one of the things that California is trying to do here is make the law narrowly tailored enough so that it can try and withstand a constitutional challenge. So I think 365 days a year political deep fake would be unconstitutional. That's not narrowly tailored. So here, California is trying to draw that line as close as possible while also preserving its interests and saying, look, there's a point to saying there is some time at which it becomes dangerous to circulate these ads.

A (18:21)

Yeah, yeah. Because I find that really interesting, you know.

B (18:24)

Yeah, it is very interesting.

A (18:26)

What is the level of, like, what is the distance from an event where you can no longer, you know, like, present the truth and therefore, you know, it causes a problem? You know?

B (18:36)

Right. And you know, why 60 days, exactly 40 days, or why not even longer?

A (18:41)

Yeah, yeah.

B (18:43)

And so that's satire and parody. But I think there's a different kind of question here, and that is if satire and parody are protected speech, but maybe perhaps some kinds of political deep fakes aren't, or they can be highly restricted, are we sure that we can tell the difference? Right.

A (19:00)

Yeah.

B (19:01)

Because that line between fake and deceptive and therefore bad and satirical maybe isn't always clear. And so that raises a problem as well. It makes the policy problem of election deep fakes a real issue. Then there are these practical problems, too. Even if you have a constitutionally drafted ban on election deepfakes, how are you going to enforce it? Michigan passed an election deepfake ban just this year, but if you violate it, it subjects the person to no more than 90 days in jail or a fine of $500. Not sure that that's enough to deter lots and lots of people from participating in these deceptive ads. Or let's imagine you're a foreign adversary. I'm not sure that $500 is going to cut it. And even if you can find the person responsible for an illegal deepfake. If that video or audio goes viral, the damage has already been done. Just trying to figure out who's that one person who started it all is really just one small part of the issue. And the very existence of deepfakes, I think, can alter our perceptions of reality. We've already seen that some famous people have tried to cast doubt on real videos by saying, well, maybe it's a deepfake. That's not really me. And that's the problem. Once deepfakes exist, how can you trust anything when you know some things are fake? And what happens when it's not clear what's real and what isn't? Or what if nobody really even cares that much? And that leads us to Springfield, Ohio. This episode is brought to you by Progressive Commercial Insurance. Business owners, meet Progressive Insurance. They make it easy to get discounts on commercial auto insurance and find coverages to grow with your business. Quote in as little as 7 minutes@progressivecommercial.com Progressive Casualty Insurance Company coverage provided and serviced by affiliated and third party insurers. Discounts and coverage selections not available in all states or situations. So let's talk about this story. Since 2020, about 20,000 Haitians have moved to Springfield to live and to work. And many of them have relied on a program that temporarily protects them from deportation. Their arrival in Springfield, which has a population of about 58,000 people, has caused some tensions in the city. But all of this became a national news story because in early September, a conspiracy theory started circulating on social media that Haitian immigrants were stealing, killing and eating pets in Springfield, Ohio. Now, just to be clear, there's never been any evidence that this is true. But that rumor kept circulating online and shared widely. And on September 9, Trump's running mate, J.D. vance, claimed on x former Twitter that people have had their pets abducted and eaten by people who shouldn't be in this country. And then there was the presidential debate on September 10 between Trump and Harris. Harris made several points during the debate that were designed to bait Trump to get under his skin. And when Harris attacked Trump for killing a bipartisan immigration bill while he was president, she continued by trolling him about his rallies. Harris said, look, he goes on and on about Hannibal Lecter and windmills. People get so bored, they start leaving his rallies. And Trump took the bait. He said, no, people don't leave my rallies. Then Trump said this, a lot of towns don't want to talk about it because they're so embarrassed by it. In Springfield, they're Eating the dogs, the people that came in, they're eating the cats, they're eating the pets of the people that live there. Trump repeated the lie. And so when a presidential candidate repeats an ugly lie like that, it has real life consequences. Since the debate, the city of Springfield has been flooded with threats of violence. They've evacuated the elementary schools. Colleges in the area canceled their classes. The city hall and county buildings in the city have been closed. Some of the Springfield hospitals went on lockdown. And again, this is a completely false story. But the reason to bring up the Springfield story is because of what J.D. vance said on September 15, after city officials came forward and said, no, no, these stories aren't true. Vance appeared on CNN and said, if I have to create stories so that the American media pays attention to the suffering of the American people, then that's what I'm going to do. That's the environment in which election deepfakes thrive, when lies are helpful and the truth doesn't matter.

A (23:58)

I literally hate that man. It's just like, it is just the most ridiculous thing in the world to say. Like, it's just I'm going to lie to further some point than I have. I mean, it's also like, this is a person who is the senator of that state. That is ridiculous that he wouldn't represent everyone in that. I'm so furious. Okay, so besides JD Vance being a morally reprehensible person without a conscience, is there anything else that we need to add to this discussion about deepfakes?

B (24:32)

Yeah, one last thing. On September 17th, California's Governor Gavin Newsom signed into law an update to its 2019 deep fake law. Now, one of the things the new law does is to expand the time when the ban applies. It's now illegal to distribute an election related deepfake 120 days before election day and now 60 days after. The law took effect immediately. So it applies to this presidential election. So remember, I brought up this idea of a Deepfake video that looks like voter ballots are being dumped in an election. Right. That could cause problems even after election Day. And the new law addressed that kind of a problem. Now, in July, Governor Newsom said one reason that he was going to sign the bill into law was because of a deep fake of Kamala Harris, which was then shared widely by Elon Musk on x. He has 198 million followers in the Deepfake. You see an altered Harris campaign ad. Kamala Harris says things like, I was selected because I'm the ultimate diversity hire. It looks and Sounds pretty realistic. It's completely fake. Now, after Newsom actually signed the updated bill into law on September 17, Musk reposted that same deepfake video of Kamala Harris. Of course, we're now less than 50 days before the election. Musk posted the video as a protest. This time, though, there was a difference in his post. This time, Musk's post contained the label parody.

A (26:12)

Okay, that's doing something. Yeah. This is a parade of horrible men we're talking about today.

B (26:23)

Well, it does raise the question of are these laws effective? Right. If you're immunized by just saying, well, it's a parody, so it doesn't matter.

A (26:31)

I mean. Yeah, but I just think that.

B (26:36)

You know what I mean, if the point is these things are dangerous and therefore we should try to discourage people circulating them, just slapping the label on seems to maybe not get to the point.

A (26:47)

Yeah, but it's still like a step in the right direction. And even the most robust of these laws, is it the creation of the deep fake that is against the law or the dissemination of, of a viral deep fake, which I feel like that is completely unenforceable.

B (27:02)

Yeah, I mean, the laws target both kinds of problems and they pose different problems. I mean, one just for the people who are in the business of doing this. People and institutions and even foreign actors, frankly, who want to influence our elections. And then people just sharing things without really taking a hard look at whether it's real or fake. And maybe sometimes not caring.

A (27:26)

Yeah, yeah. Well, I mean, this stuff is super fascinating and just like swimming in this sea of misinformation is becoming just like a full time job, like navigating where you are in it. And I kind of don't know what is going to happen. You know, like, it just seems. It feels completely new.

B (27:48)

It does. And I think one of the things that we do here is talk about how our constitutional traditions can protect us or kind of reinforce our ideals. But with deep fakes, I'm not sure that the machinery is adequate to the task.

A (28:05)

Yeah, completely. It's kind of the equivalent of the Second Amendment. It's just like I personally am of the conviction that the second Amendment wouldn't read as it does if the people writing it were aware of automatic weapons and weapons of mass destruction in the way that they are. And the same goes for the ways that misinformation is spread. I mean, it's just, it's so heightened. It's so much more intense. It so much has much more ability to fool reasonable people that there would be some kind of different nuanced thought about the First Amendment and freedom of the press in that scenario as well.

B (28:50)

Yeah, I think that's right. And part of that is the technological environment that has changed what it means to be part of political debate, the ways in which a lie can spread so quickly. But also even that comment advances, which is, who really cares about the truth at all? Right. I'll make up things if I need to.

A (29:12)

Yeah. This is like the Constitution doesn't like, have a remedy against useless assholes, you know, I mean, just like it's the worst. It can't. I mean, that's what we've learned. This is like the project of this, of this podcast is just like, you know, sometimes it's like Constitution just like it is not inoculated against someone who just lies, who is just out for themselves. It is really like its sort of basis and how it functions is kind of like the goodwill and sincerity of the people in charge. And when you don't have that, it really just breaks down so quickly. And it's stunning to me that, you know, to sort of like watch those cracks, you know, form in real time.

B (30:01)

So anyway, yeah, and it also goes to our, you know, are we looking at speech in the right way in 2024? I know that's, you know, thing that people talk about sometimes. And I should add that the video that was posted and then reposted by Musk, the deep fake of Kamala Harris, the creator of that video after the law was signed by Gavin Newsom just this week, has already filed a lawsuit in federal court saying that the law violates his First Amendment rights.

A (30:33)

Well, I'm going to practice my First Amendment right and saying that I hope that person loses that case badly. Is this also bad?

B (30:45)

Yeah, I mean, I'm not sure that the California law is going to accomplish the goals that it has, but we'll see.

A (30:52)

No, but sometimes, I mean, I understand that you need laws to function, but I also understand laws as a statement of values. And then you go through the courts to sort of test and, you know, explain those, you know, like, you know, put that into real world action. Like, like it's, it's sort of like a prototyping process, but, like, as a statement of values. I think that that law is. It represents my values as a Californian.

B (31:21)

I think that's right. And I think part of the challenge is going to be very often these kinds of laws are trying to make political points by saying we're addressing the problem, but they do raise all kinds of questions like, you know, parity is not defined because you kind of can't define parity. Sure. But the nature of parody is pretty different with a deepfake. Right. No one would have read the Hustler ad of Jerry Falwell and thought, well, did he really say that? Nobody thought that. Right. But with deepfakes, there is that moment, or maybe you never realize that it's completely made up.

A (31:51)

And that was the basis of the defense in the Larry Flynt case, was that if Jerry Falwell is this upstanding pillar of the community, no one reasonable believed it. And therefore, then, therefore it wasn't something that that mattered. And it seems like the whole intention of a deepfake is to fool people. Like, it's like if we're going to assign the idea of malice to these things, then you could maybe also assign the idea of fooling people as sort of like, as a subset of malice.

B (32:22)

Yep. And I think, you know, the tools are so good that it really is mostly deception and less parody.

A (32:29)

That's right. Well, this is fascinating stuff. Thank you so much, Elizabeth. I appreciate it.

B (32:33)

Thanks, Roman.

A (32:35)

This show is produced by Elizabeth Jo, Isabel angel and me, Roman Mars. It's a mix by Hazik Ben, Ahmad Fareed. Our executive producer is Kathy Tu. You can find us online@learnconlaw.com all the Music in what Roman Mars can learn about Con Law is provided by Doomtree Records, the Midwest hip hop collective. You can find out more about Doomtree Records, get merch and learn about who's on tour@doomtree.net we are part of the Stitcher and SiriusXM podcast family. Can lighter structures really be stronger? Yes, they can. ExxonMobil is helping advance American industrial innovation with Proxima Systems, enabling a lighter and stronger alternative to traditional rebar while lowering greenhouse gas emissions to help build a more efficient, efficient construction industry with sustainability in mind. ExxonMobil. Let's deliver.