1y ago

Teen deepfake victim pushes for federal law targeting AI-generated explicit content

www.nbcnews.com Teen deepfake victim pushes for federal law targeting AI-generated explicit content

Rep. Joe Morelle, D.-N.Y., appeared with a New Jersey high school victim of nonconsensual sexually explicit deepfakes to discuss a bill stalled in the House.

News @lemmy.world

www.nbcnews.com /tech/tech-news/deepfake-law-ai-new-jersey-high-school-teen-image-porn-rcna133706

40 6

116 comments

Think of the children being used to push an agenda that helps the very wealthy? Well I'll be, what a totally new and not at all predictable move.
Ban all ai that aren't owned by rich people, make open source impossible, restrict everything that might allow regular people to compete with the corporations - only then will you children be safe!
- I'm as suspicious of "think of the children" stuff as anyone here but I don't see how we are fighting for the rights of the people by defending non-consensual deepfake porn impersonation, of children or anyone.
  If someone makes deepfake porn of my little cousin or Emma Watson, there's no scenario where this isn't a shitty thing to do to a person, and I don't see how the masses are being oppressed by this being banned. What, do we need to deepfake Joe Biden getting it on to protest against the government?
  Not only the harassment of being subjected to something like this seems horrible, it's reasonable to say that people ought to have rights over their own likeness, no? It's not even a matter of journalistic interest because it's something completely made-up.
  
  We're not talking about whether we should make fakes. We're talking about whether people who do, should be prosecuted - IE physically overpowered by police officers, restrained with handcuffs, and locked up in a prison cell. Some empathy?
  If some classmate of your little cousin makes a fake, should the police come and drag them out of school and throw them in prison? You think that would help?
  Realistically, it's as likely to happen as prosecution of kids who "get into fights" for assault. Kids tell mean lies about each other but that is not resolved in civil suits over defamation. Even between adults, that's not the usual thing.
  Civil suits under this bill would be mainly targeted against internet services, because they have the money. And it would largely be used over celebrity fakes. That's the overwhelming part of fakes out there and they have the money to splurge on suing people who can't pay. It would be wealthy, powerful people using it against horny teens.
  Also, this bill is so ripe for industrial abuse. Insert a risqué scene in a movie, and suddenly "pirates" can be prosecuted under this.
  
  The issue is there really is no way to stop it unless you make ai illegal. The cat is already out of the bag. The models and hardware are getting better and faster and cheaper.
  How do you suppose you enforce a law like this when people stop even sharing the photos they create, maybe don't even save them themselves, because it's so easy and instant to create more when you want to see them. "Put her face on her body in this position", bam, instant album of photos to jerk off to, then delete them. That's how good and how available these models are getting.
  How do you think restrictions on this should, or could, be enforced?
  
  Mr. Vulpine, what you've just said is one of the most insanely idiotic things I have ever heard. At no point in your rambling, incoherent response were you even close to anything that could be considered a rational thought. Everyone in this chat room is now dumber for having listened to it. I award you no points, and may God have mercy on your soul.
- deleted
- You kind of have to be rich in order to run these image generation AIs. The RTX 4090 TI isn't cheap.
  
  You kind of have to be rich in order to run these image generation AIs. The RTX 4090 TI isn't cheap.
  Any iPhone or iPad on the current version of iOS can run Stable Diffusion locally with the (free) Draw Things app.
  Hell, if you’re willing to run on the CPU instead of the graphics card (which takes much longer) you can get Stable Diffusion working on pretty much any PC. And honestly any semi-recent nVidia card will have drivers to run it.
  What’s more, there are free sites for SD image generation.
  Image generation isn’t expensive, and it gets cheaper and cheaper every year.
  
  The NVIDIA 1660 ti is perfectly adequate and not only the property of rich people.

Creating fake child porn of real people using things like Photoshop is already illegal in the US, I don't see why new laws are required?
- Well those laws clearly don't work. So we should make new laws! Ones that DEFINITELY WILL work! And if they don't, well I guess we just need more laws until we find ones that do.
  
  Since we need a rule explicitly for AI related cases, even though it's already covered by others, lets ensure that we also make a 100 page law for if the material is explicitly made in Photoshop, and also another 80 pages if it was made in Gimp. If you use MS Paint to do it, we need a special 200 page law that makes the punishment even harsher, because damn you got skillz and need to be punished more.
  
  deleted
- This is not at all about protecting children. That's just manipulation. In truth, kids are more likely to prosecuted than protected by this bill.
  There are already laws that could be used against teen bullies but it's rarely done. (IMHO it would create more harm than good, anyway.)
  This is part of an effort to turn the likenesses of people into intellectual property. Basically, it is about more money for the rich and famous.
  This bill would even apply to anyone who shares a movie with a sex scene in it. It's enough that the "depiction" is "realistic" and "created or altered using digital manipulation". Pretty much any photo nowadays, and certainly any movie, can be said to "altered using digital manipulation". There's no mention of age, deception, AI, or anything that the PR bullshit suggests.
- Regulatory capture. OpenAI wants to kick down the ladder

Just wait until them tech savvy folks in Congress try to understand the difference between 'deepfakes' in the sense of pasting a new face on existing footage and whole cloth generative AI creating the entire scene, and then someone tells them that the latter is derived from multiple existing media sources. Gonna be some smoke pouring out of their ears like in the cartoons trying to slice up all the specifics.
- Just tell them the 2nd amendment guaranteed your rights to Gaussian Splatting.

I really wonder whether this is the right move.
This girl, and many others, are victims and I don't want to diminish that, but I for better or worse I just don't see how legislation can resolve this.
Surely deepfakes will be just different enough to the subject to create reasonable doubt that it depicts the subject.
I wonder whether, as deep fakes become commonplace, people might be more willing to just ignore it like any other form of trolling.
- It’s not trolling it’s bullying. You need to think beyond this being about “porn”. This is a reputational attack that makes the victim more likely to be further victimised via date rape, stalking, murder. These things already happen based on rumours, deepfakes images/videos will only make it worse. The other problem is that it’s almost impossible to erase once it’s on the internet, so the victim will likely never be free of the trauma or danger as the images/videos resurface.
  
  Trolling / bullying is just semantics, which I don't think will help us very much.
  I think the heightened risk of other crimes is... dubious. Is that conjecture?
  Your position seems to be framed in the reality of several years ago, where if you saw a compromising video of someone it was likely real, while in 2024 the opposite is true.
  Were headed towards a reality where someone can say "assistant, show me a deepfake of a fictitious person who looks a bit like that waitress at the Cafe getting double teamed by two black guys". I don't claim to know all the ethical considerations, but I do think that changing social norms are part of the picture.
  I don't have any authority to assert when anyone else should feel victimised. All I know is that in my own personal case, a few years ago I would've felt absolutely humiliated if someone saw a compromising video of me, but with the advent of deep fakes I just wouldn't care very much. If someone claimed to have seen it I would ask them why they were watching it, and why in the world they would want to tell me about their proclivities.
- I hope it won't overregulate technology itself but instead would be ruled by already existing means about defaming people and taking photoes without their consent, sharing them. Plus, if she's a teen, it's a production of CSAM. This person had an illegal intent, just used a new tool not unlike others, just more efficient.
- Surely deepfakes will be just different enough to the subject to create reasonable doubt that it depicts the subject.
  That's a major assumption. Do people really think a school board will really consider that when a student creates a fake Only Fans of a teacher? A random University or Company doesn't even give reason for denying an application when they see any form of online nudity? People are lazy as fuck and will just move on to the next candidate or let someone go to save their own image rather than that off the victim.
  
  My point is, when it becomes as easy to generate deepfakes as it is to order your groceries, the question will become "why is the university searching for deepfakes of everyone"
- I think you're right if the goal is to stop them all together.
  But what we can do is stop people from sending them around and saying that it's true/actually the person.
  Once they've turned it from a art project into a weapon, it should have similar consequences to "revenge porn."
  
  I would think this would be covered by libel, slander, defamation type laws. The crime is basically lying about a persons actions and character.
  
  I don't know how strong the laws are on the topic but I feel this falls under harassment or libel. In most cases this will cause emotional distress and harm to a person's reputation. If you're trying to show off your AI skills you can use a subject that isn't real or depict a real person wearing clothes. This is clearly an attack in my mind.
- My dude there are people out there thinking they're in a relationship with Johnny fucking Depp because some Nigerian scammer sent them five badly photoshopped pictures. Step out of your bubble, maybe. This shit isn't easy to spot for the vaaaaaast majority of people and why would this lie with the victim to sort of clear their name or hope that idiots realize it's fake?
  Especially with and around teenagers who can barely think further than their next meal?
  Good lord.
  
  WDYM "step out of your bubble"?
  It's not a question of being able to detect whether or not a video is fake. When deepfakes become so prevalent that everyone's grandma understands that they're prevalent, it won't matter whether you can identify the video as fake.
- Photos of a person can vary in subtle ways too, perhaps as a person ages or even just changes their makeup or something. It's not valid to require everything to be perfectly clear-cut in some objective way.
  Life is subjective, which is why courts always try to take the mental state of the accused into account, things like whether malice was present, whether the accused was in a rational state of mind, etc. This is why we have first and second degree murder as two different things.
- I think it doesn't go far enough. Straight up, no one should be permitted to create or transmit the likeness of anyone [prior to, say, 20 years following their death] without their explicit, written permission. Make the fine $1,000,000 or 10% of the offender's net worth, whichever is greater; same penalty and corporate revocation for any corporation involved. Everyone involved from the prompt writer to the work-for-hire people should be liable for the full penalty. I can't think of a valid, non-entertainment (parody/humor), reason for non-consensual impersonation - and using it for humor or parody is a slippery slope to propaganda weaponization. There is no baby in this tub of bathwater.
  
  I'm not sure this is practically possible.
  A $1m penalty is more or less instant bankruptcy for 99% of the population. It's probably not much of a deterrent for, say an 18 year old. In my jurisdiction I don't think there are criminal penalties higher than a few thousand dollaridoos. It doesn't matter whether you think this act is so aggregious that it deserves a penalty 1000 time higher than any other, my point is that it would be unenforceable ineffective.
  Secondly, how do you determine whether an image is someone's likeness? Create any random image and surely it will look like someone, but that doesn't mean that creating that image violates that someone.
  
  Yeah, just like the FBI warnings on VHS tapes about massive fines and jail time stopped us from copying them in the 80s and 90s...
  
  no one should be permitted to create or transmit the likeness of anyone [prior to, say, 20 years following their death] without their explicit, written permission.
  I dig the sentiment. I do. And If this were my own fantasy world, I'd agree. But unfortunately, we don't live in the timeline where that is considered even close to reasonable.

FOSTA is still in effect and still causing harm to sex workers while actually protecting human traffickers from investigation (leaving victims stuck as captive labor / sex slaves for longer). And it'd still regarded by our federal legislators as a win, since they don't know any better and can still spin it as a win.
I don't believe our legislators can actually write a bill that won't be used by the federal Department of Justice merely to funnel kids for the sake of filling prison cells with warm bodies.
We've already seen DoJ's unnuanced approach to teen sexting which convicts teens engaging in normal romantic intercourse as professional producers of CSAM.
Its just more fuel for the US prison industrial complex. It is going to heavily affect impoverished kids caught in the crossfire while kids in richer families will get the Brock Turner treatment.
This bill is wholly for political points and has nothing to do with serving the public or addressing disruption due to new technology.
Until we reform or even abolish the law enforcement state, anything we criminalize will be repurposed to target poor and minorities and lock them up in unconscionable conditions.

There are already laws against creating false content about people, so adding more laws isn't going to make the previous laws more or less valid and it's only going to waste time and money.
Of course it's being pushed by a "teen" since this teen clearly doesn't have any understanding of the issues at hand, the technology at hand nor the laws that already exist to help them with this issue.
It was up to the adults around this teen to help her navigate the issue and instead the incompetent pieces of worthless shit choose to push a new bill against AI rather than use the current legal framework that exists to actually help this girl.
Anything to abuse a child or teens situation for their political gain. Worthless trash.
- deleted
  
  Pretending abortion is in the same realm as AI tool abuse is ridiculous and completely dishonest.
  .
  Scare quotes? Are you for real? Since you don't understand, quotes denote emphasis or specificity, not emotion. "Teen", the emphasis here, was used because teens are barely educated in legal matters and it's their responsibility to seek assistance, not push legislation.
  .
  If the adults in that area had actually responded properly, there wouldn't be an article about a new bill against AI, instead there'd be an article about defamation and debasement of a minor (likely by another minor, but that doesn't mean the other minors parents are infallible in this situation, you are either parents or you're not, if you are, you're responsible for your brats actions, if you're not, then the state will take the child and destroy them mentally, and likely physically, for your failures as a parent).
  .
  Not a single thing I said dismissed this, and I would go even further to say anyone pursuing the AI angle is the one dismissing this, especially given the laws that already exist that can be used to assist and protect the teen in question.
  .
  We've had adobe premier for more than 2 decades before generative AI and all of the issues surrounding deepfakes then cover every issue with generative AI now and wasting peoples time and money on this does not help improve the situation, instead it makes a mockery of the victim for the purposes of pushing an agenda.
- It's being pushed by someone who was a victim of deep fake ai porn, so I think they understand the issues at hand just fine, you don't have to agree with her, but don't be a patronizing asshole about it.
  
  Thanks for not actually reading my comment and making it clear to everyone who did that you're either illiterate or a dishonest asshole.
  
  Disagreeing strongly isn't patronizing.

This is just the tip of the iceberg of the threat AI poses to human social structures. We have yet to appreciate the gravity of what these new technologies enable. It's incredibly dangerous yet equally naive to think that AI-generated porn laws will keep us safe.
Firstly, the cat's out of the bag. We can ban the technology or its misuse all we like, but can we really practically stop people from computing mathematical functions? Legal or not, generative AI can and will be used to generate content that hurts people. We need better planning for identifying, authenticating, and responding when this misuse happens.
Secondly, we have an already huge, huge problem with fake news and disinformation. What is such a law for this special case of AI porn going to do for our inability to address harmful content?
It's a shame, but it strikes me as more feel-good than actually doing something effective.
- People said the same thing when, after the printing press, there was rampant plagiarism and reverse-plagiarism (attributing words to someone who never said them).
  After a period of epistemic chaos, the result was several decades of chartered monopoly and government censorship to get it under control.
  I hope we won't need heavy-handed regulation this time around. But that will only happen if we learn from history. We need to get this under control now, while we have the chance to start a framework for protecting our fellow human beings from harm. Complaining that it's hard is not an excuse for doing nothing.
  
  deleted
- A challenge is that generally the same technology to detect ai content can be used to improve it. It's gonna be an arms race
- Just ban any instance of generative unconsentual generative nudity, AI or not.
  
  I just drew a stick figure of Johnny Depp and it's indeed naked. And I'd do it again, too.

More laws that will only affect people who follow laws
- I mean, they do usually effect the people that break them and go to prison too.
  
  You mean the poor and disadvantaged?
- It's like how DRM only hurts people who purchase content legally.
  It's been very illegal to pirate games for decades, and still pirated content is quite common in the wild. What's banning it (defamation) harder going to practically achieve?
  
  Literally nothing, but morons will pretend otherwise in order to waste money and time because they can't think of any other way to engage the problem.
- Isn't that... all laws?
- Might as well not make any laws then.
- How? How will this only effect people who follow laws? If you aren't making ai porn of real underage girls how would this affect you, a law abiding citizen?
- deleted

I would have thought that deepfakes are defamation per se. The push to criminalize this is quite the break with American first amendment traditions.
If I understand correctly, this would put any image hoster, including Lemmy, in hot water because 230 immunity is only for civil suits and not federal criminal prosecution.
- the text of the bill exempts service providers from any liabilities as long as they make a good faith attempt to remove it as soon as they are aware of its existence. So if someone makes AI generated revenge porn on your instance as long as you take it down when notified, you want be in trouble.
- It's not, people know it's a deepfake most of the time and don't claim it's real
  
  It might also be harassment.
  If it's not defamation or harassment, then I'm not sure what the problem is. As broad as this is, it looks unconstitutional to me.

There is no world where a law aimed at this type of thing will ever be used for its intended purpose.

How different is photoshopped fakes from AI fakes? Are we going to try to bad that too?
ETA: ban that too. Thx phone kb.
- What does the method matter? If the result is an artifact that is convincing enough for the average person to believe that the subject knowingly posed for sex acts that never occurred, the personal experience and social stigma is traumatizing no matter how it was made.
  As the sociologist Brooke Harrington puts it, if there was an E = mc² of social science, it would be SD > PD, “social death is more frightening than physical death.”
  
  What does the method matter?
  That's my point. If we're going to ban AI fakes should we then ban ALL fakes? Where do we draw the line and how do we do that without limiting free speech? I'm not sure it is possible.
  And the days of believing everything you see are over but most don't know it yet.

This is the best summary I could come up with:
A teenage victim of nonconsensual sexually explicit deepfakes joined Rep. Joe Morelle, D-N.Y., on Tuesday to advocate for a bipartisan bill that would criminalize sharing such material at the federal level.
In addition to criminalizing the nonconsensual sharing of sexually explicit deepfakes, the measure would also create a right of private action for victims to be able to sue creators and distributors of the material while remaining anonymous.
Mani said her school administration told her on Oct. 20 that male classmates had created and shared sexually explicit deepfakes of her and more than 30 other girls.
After he heard about what happened at Mani’s high school, which is in his hometown, Rep. Tom Kean, R.-N.J., became the first Republican co-sponsor of Morelle’s bill.
The lack of legislative movement around deepfakes has raised concerns about the technology’s potential to disrupt the 2024 election cycle.
A legal expert who specializes in nonconsensual intimate imagery, Mary Anne Franks, who Morelle said helped inform the bill, said deepfakes have already targeted female politicians.
The original article contains 457 words, the summary contains 169 words. Saved 63%. I'm a bot and I'm open source!

I feel we are in need of a societal shift here, just like another commenter said about the printing press. When that first came out, the pushback was from the worry that the words would be attributed to someone who never said them (reverse plaigerism). The societal adjustment to this was the universal doubt that anyone said that thing without proof.
For generative AI, when it becomes widespread, photos will be generateable for literally everyone, not just minors but every person with photos online. It will be a societal shift; images will be assumed to be AI generated, making any guilt or shame about a nude photo existing obselete.
Just a matter of time so may as well start now!
- I agree that this is probably the inevitable end result of the proliferation of the technology. The journey society is going to have to take to get to that point is going to be pretty uncomfortable though I think.

If (as it seems) the point is not impersonation but damage to the person's honor/image, where exactly is the line?
If realism is the determining factor, what about a hyperrealistic human work? And if it is under human interpretation how realistic it should be, could a sketch be included?
- A sketch would probably not convince anyone that the subject consensually participated in sex acts that never occurred.
- Good question. The bill doesn't define realistic. There's no condition that it should fool anyone.
  This definitely goes beyond AI and includes photoshops, 3d renders and any other digital art. I think it would also include hand drawn images, once they are digitized, EG by photographing them on a phone. Always provided that the depictions are in some way "realistic".

116 comments