Is It Legal to Swap Someone’s Face into Porn Without Consent?
theverge.com
theverge.com
How then do paparazzi get away with photos of celebrities?
Almost all pictures of celebrities are in or from public areas where there is no right of privacy. (If a celebrity really didn't want to be photographed naked, they wouldn't undress in front of a public-facing window. There are decades of court cases on this point.) The pictures that aren't, i.e., Erin Andrews, resulted in massive court victories for the victims.
The appropriation cases are much rarer, as it generally applies to impersonation cases. It's usually in context of name appropriation, as most image appropriation cases thus far are pursued under different tort theories usually involving commercial harm. (Though since Catfish, image appropriation invasion of privacy cases are becoming more common.)
Could you explain that a bit more? I would be very shocked if something as subjective and vague as "publicizing a person in a false light" would be an invasion of privacy worthy of first amendment exemption.
Generally, any person who pursues a defamation claim over a deepfake would likely include this claim if they can't meet the burden for defamation.
(Please don't try to pick that apart as I'm only trying to provide a very-high-level-in-a-nutshell general overview.)
† edit: I'm hedging here because I'm unsure about the exact intersection.
It's not clear to me where Goldman says anything that disagrees with that. Are you sure it was something Goldman said that you are disagreeing with, or was it the quote "You can’t sue someone for exposing the intimate details of your life when it’s not your life they’re exposing"?
If the latter, that is not Goldman. That is quoting the Wired article that the submitted article cites.
This deepfake situation, and a similar situation that will occur by synthesizing fake speech in people's own voices, is unstoppable. All the complaining in the world won't stop it, because the barrier to entry will soon be trivial. What is the proposed response? Criminalize running certain kinds of neural nets? Or social media shaming anyone who publishes any fake (that's detectable)?
We all just need to get used to the fact that pretty soon audio and video recordings of people won't be verifiable without careful checking and comparing, and eventually maybe not even then.
Look at the bright side: someone can have a recording of you saying something crude and disrespectful and describing sexual assault, and if you're running for president you won't even have to admit it was you anymore, because soon anyone will be able to fake that kind of thing. Video's more difficult, but that too might happen.
Honestly this scenario is so terrifying to me that I almost hate to even discuss it. What happens when people can make convincing fakes like that of religious figures and weaponize it to incite violence or install authoritarian governments?
How much more powerful and devastating would calls for genocide be, if driven by deepfake videos? Especially in parts of the world where people wouldn't even be a little bit skeptical of something outrageous that they saw with their own eyes.
Just look at the current political climate in the US. How much more polarized and volatile would things be if someone could fake videos showing politicians abusing children or planning a violent 'government takeover'.
If/when this tech gets good enough, it will absolutely be disruptive to civilization.
No, it really won't, unless it involves them personally. If a celebrity or politician denies it was them in "that video," most people will believe whatever conforms to their biases about that person.
You do realize it was already possible to do this? What's new here is not the capability to create a convincing fake video, but the deep learning/AI aspect of it.
>How much more powerful and devastating would calls for genocide be, if driven by deepfake videos? Especially in parts of the world where people wouldn't even be a little bit skeptical of something outrageous that they saw with their own eyes.
>Just look at the current political climate in the US. How much more polarized and volatile would things be if someone could fake videos showing politicians abusing children or planning a violent 'government takeover'.
Probably not much. There aren't a lot of people out there who would commit mass murder or political violence if only there were convincing documentary evidence, but who wouldn't otherwise.
And yet here were are, for the most part people believe what they see in reputable newspapers.
With the expansion of fake photos and videos, it will just come back to trustworthiness.
Do you trust the source that has provided the image?
https://img.etsystatic.com/il/c435d0/942014524/il_570xN.9420...
If it’s your face in the image you have likeness rights in the image. If a photographer takes your photo and uses your image in a way that makes you and your face significant in the image, they need to have a model assignment agreement signed by you on file or they are setting themselves up for a lawsuit and a world of pain in the event the image goes big. Yes, there are exceptions to the likeness rights laws, but deepfake swaps are unlikely to fit into those exceptions (and yes, California law in particular makes clear those rights extend to dead celebrities through their families or the people their families have transferred the rights to, and it’s very hard to have your image go big and not have it go to California).
This is one of the things that makes the Flickr Creative Commons flag so dangerous. The photographer is saying “you can use this commercially” but 99.9% of images on Flickr are snapshots taken by amateurs with no rights assignments from the subjects. There have been and continue to be lawsuits where the photographer is found liable for the commercial use of their snapshots, without their knowledge, because they flagged it as CC and someone did and the photo subject got upset and sued.
Likeness rights (aka personality rights aka right of publicity) aren't all that simple or clear, are rapidly evolving with different forces pulling them in different directions, vary considerably between jurisdictions, etc., and, because they are rapidly evolving and inconsistent at the state level, many aspects of them may not be well-tested against overriding Constitutional provisions (First Amendment concerns.)
Most personality rights / right of publicity regimes apply to promotional use and a subset of commercial use (California, noted frequently for having a generous regime for such rights, still explicitly notes that content merely being monetized with advertising is not necessarily enough to qualify for protection.)
I don't think that covers most deepfakes.
But frankly, I'm okay with it. I think these celebrity/trump/ajit images are clearly protected under the first amendment. And I'm glad they are.
Of course there is a difference between something having more sexual than political gratification. But that seems like an odd and murky line to draw here.
My question would if there would be a diff btwn public person (aka famous, known) and run of the mill hoi polloi. Famous personalities in some circumstances have less recourse in terms of privacy.
Should the (many!) Elvis impersonators need a licence from his estate?
Impersonating others isn't a new issue, we already have a quite settled body of law about that and centuries of weird edge cases. I don't see a radical difference brought in by technology here. IMHO the treatment of making face-swapped porn videos (or distributing them, which is a quite different case) should be the same as the currently established treatment of painting (or distributing) a realistic painting of someone you know in a pornographic pose - whatever that treatment is.
In any case, this is more narrow in that its focus is on pornographic representations of regular people (a la revenge porn).
Right of publicity / personality rights are actually a fairly actively evolving area of law (and one with considerable variation between jurisdictions, even with the US), not an area of settled consensus.
Actually, the way the law currently works...they might (to avoid one or more tort cases for invasion of privacy, defamation, and/or publicity rights). Studios will rename characters in biopics if they can't secure rights to the person the character was based on. Like all things in the law, it is a matter of context. A biopic or other media "based on a true story" may require the permission of the people portrayed in it. A satire generally would not...The law is complicated like that.
Should the (many!) Elvis impersonators need a license from his estate?
Elvis is dead...The publicity rights of the dead are generally minimal and the invasion of privacy rights nonexistent...Also, the Elvis estate is generally okay with impersonators because for the most part they don't harm (i.e., defame) his image. They have gone after impersonators who they believed misappropriated Elvis' image for defamatory purposes.
IMHO the treatment of making face-swapped porn videos (or distributing them, which is a quite different case) should be the same as the currently established treatment of painting (or distributing) a realistic painting of someone you know in a pornographic pose - whatever that treatment is.
Under the current law, that is several torts: Invasion of privacy, publicity rights, and defamation...
What circumstances would require a film production to ask permission of a real person before an actor could portray him? E.g. The film "The Social Network" didn't ask Mark Z for permission to portray him in an unflattering way:
https://movies.stackexchange.com/questions/2108/how-can-a-st...
Invasion of privacy did not apply because he was a public figure and held to the higher standard for IOP claims. Pretty much everything in the movie has already been revealed before, so there isn't anything private that was revealed in the movie itself.
Mark Z could have had a case for publicity rights. The only problem is...what damages did he suffer? It's likely that his legal team realized that they could only ever win a relative pittance in damages but not block the distribution of the movie, and suing would just trigger the Streisand effect, so they just did nothing. (His damages would likely have been based on the amounts paid to other celebrities for their life stories...usually in the range of $250k to $5 million, a pittance for a billionaire.)
I would certainly bloody hope they need permission. You have the usual caveats regarding the public interest for things, but I mean, one could argue a director should at least have one face to face conversation with someone they're including in a movie.
> Should the (many!) Elvis impersonators need a licence from his estate?
I think it's super creepy how widespread and how effectively public domain Elvis' image, act, and etc. is. He's been impersonated, lampooned, parodied, etc. so many times over. I mean you can't deny his cultural impact, but I don't know. I find it unsettling.
> Impersonating others isn't a new issue, we already have a quite settled body of law about that and centuries of weird edge cases. I don't see a radical difference brought in by technology here. IMHO the treatment of making face-swapped porn videos (or distributing them, which is a quite different case) should be the same as the currently established treatment of painting (or distributing) a realistic painting of someone you know in a pornographic pose - whatever that treatment is.
More to the point: I'm not talking about impersonation, I'm talking about the use of one's likeness for personal, commercial, or illicit purpose without that individual's knowledge or consent, and frankly I find HN's overall response to this a little alarming. It seems (though this admittedly small and totally not even remotely scientific sample) that the tech-inclined don't seem to care much for privacy rights when it stops them from doing objectively cool things.
While the details may need tweaked, many jurisdictions already have “something akin to copyright to cover our own faces and identity in general”, under the name “personality rights” or “right of publicity”.
More often, “in our increasingly non-local economic systems”, laws (especially civil laws, where the complaining party chooses whether and where to pursue violations) that vary by area are effectively enforced at the level of the most favorable (to the complaining party) regime available, rather than being unenforced.
The reason social networks and other apps can get away with it is because you agree to let them use images of you posted to the network for commercial purposes as part of the Terms of Service you agree to when you use the service. In the absence of that provision in a ToS, you would have a very valid claim for appropriation of image if a social network used a picture of you.
These TOSs do not extend to other pictures of you. Thus, for example, Facebook couldn't run a commercial with a picture of you that was posted to Twitter but not Facebook.
As to your advertising idea, that's interesting but also brings up issues don't you think? Especially now with better facial recognition technology. For example, walking down a street, a camera scans your face, pull up a bunch of data on you (say you've been looking for a kayak lately), then in a nearby billboard or hologram display area, they show you images of yourself and your family kayaking down the Colorado or in an ocean.
What's interesting in movies/tv is that as the technology matures, we may never see some of our stars actually age or even what they truly look like. A lot of time and money is already spent on beauty work so it's not too big a leap to see this being mixed in with that.
Interesting you should mention facial recognition, because that is what I turned the technology into - or reverted it back into. My 'digital double' creation for automated actor replacement involves neural net trained 3D reconstruction of faces. It originally came from pose correction for facial recognition. Being an animation/VFX, guy I licensed the 3D reconstruction tech from an FR company.
After my startup efforts failed, the FR company hired me because I'd done the largest scale use of their technology at the time. Now I create mesh network appliances of their tech, combined with my production tech from my feature animation days, to create apps for dynamic FR security perimeters. The interesting thing is, the frame processing I do could easily be fed into a 'deep fake' system to produce crazy fake media pretty much automatically. We track faces frame to frame and project a 3D polygon mesh matching their facial expression. Need more?
As much as I read about new tech stuff, I had no idea how far things were along until all the deepfake porn headlines hit the news the past few weeks. I guess amidst all the hand-wringing over Carrie Fisher and Grand Moff Tarkin's "looks" in Rogue One it never really struck me that that stuff had 'arrived'.
There are some fun grey areas where you have someone doing porn who actually looks like a celebrity.
The good news is that you should be able to easily detect a face swap video, by comparing the face against public images that were taken of the celebrity and comparing the rest of the video with porn videos (like Youtube's ContentID).
There a lot of stuff that people should rightly be able to do without an affected person's consent (such as leaving a negative review of their work), but having framing that centers on "consent" obscures that.
So, if such a law would be implemented, it would be much closer to DRM laws, which allows data authors and content right owners to dictate how you can process their data, even if they give it to you. In one case, it's copying the data to another format, or may be running a decompiler on it; in another, it's using it in deep learning. Either way, it's processing.
Releasing new bits would be invasion of privacy and defamation.
"Processing" existing bits would be defamation, appropriation, and possibly invasion of privacy based on the resulting images, though the IOP claims would likely be add-on claims to a defamation or appropriation action.
Either way, it's a tort.
What if the statement or image was labeled as a fake from its inception, not just "after the fact"?
By your logic, this statement is defamatory/libel:
"This statement is false: Barack Obama killed a man just to watch him die and also cheats at cards."
The apple is the original scenario: a deep fake with an accompanying and severable statement, not part of the deep fake, that the image/video is fake. (The severability of the statement matters. If you want to know why, get a Westlaw subscription and spend a few weeks reading defamation case law.)
The orange is the statement about BO. The claim of falsity is intrinsically tied (structurally and contextually) to the false statement accompanying it; they are a single contextual unit. Moreover, the claim of falsity precedes the allegedly defamatory statement to which it is unseverably tied so it is clear at the time of publication that the statement is intended to be false. And let's not even get into the heightened standard that applies defamation claims for public figures...
> The claim of falsity is intrinsically tied (structurally and contextually) to the false statement accompanying it; they are a single contextual unit.
I don't see how that can't or couldn't apply to a deepfake: the creator could name the file making clear it was a fake, or put a caption in corner of the video stating so.
The "this statement is false" in my example seems just as separable as that. Someone could edit it out or fail to quote it, etc.
Some of your other comments are about trivial issues: the grammar could be trivially changed to put the clause noting falsity at the end. I also used a public figure as an example, because the name came to mind quickly, but you might as well substitute a normal person's name for Obama's.
You can certainly argue that the defendant didn't intend to defame someone and included a statement that the image was fake. But as I said, that statement would not eliminate your liability for the tort; it would merely mitigate the damages.
Tort law is complicated and isn't something you can just pick up by browsing the internet for a few minuets. I've handled these types of cases before. Have you?
Of course. OTOH, if you build it into the image/video, that's trivial to assure for each publication you are involved in, and subsequent publication by some other third party of some subsequent modified image that removed the statement of falsity is no different than if someone republishes a modified version for your statement that someone is not a child molester which omits the “not”.
> Tort law is complicated
(9_9)
> and isn't something you can just pick up by browsing the internet for a few minuets.
Yeah, I've never really questioned the decision I made to actually buy and study the required texts, etc., when I took Torts, rather than just spending a few minutes studying on the internet, so I really don't need anyone to validate that for me.
(Now, if we were talking about false light, in the jurisdictions which recognize it, rather than defamation, this would be a trickier issue, I think.)
That's not what I was referring to.
People who are publishing these videos are at the same time publicly say that these are works of fiction and fake. Not sometime later; right at the same time. They publish it in the reddit that has "fake" in the name. They often put "fake" in the title. They sometimes watermark it with a word "fake".
How can this be defamation or libel?
By posting it to reddit or online, the deepfaker should have expected that it would be further disseminated (and that is the standard they will be held to, even if they didn't actually expect that). In order for statement of falsity to be a defense to defamtion, they would have to take steps to reasonably ensure that the statement accompanies the further dissemination of the video/image. This means intrinsically tying it to the fake content. A filename wouldn't be enough; they would need to embed it into the video itself, at the points where the images are faked, into order to avoid liability for defamation. Anything less than that and every jury in the country would find them liable for defamation.
Tort law is complicated. The deepfaker of Daisy Ridley is in for a world of hurt if she chooses to pursue a defamation claim. On the other hand, the news sites that posted clips from the deepfake are not.That's an entirely separate discussion...
Would a good watermark suffice, for example?
I'd expect the law to catch up soon, and that to happen after the issue is made more common once Reddit's non-consensual deepfakes goes more viral than it already has.
Edit: replying to the flagged response below, which I disagree with but which I think deserves a reply:
By immoral, I mean hurting other people without consent. I do not believe any activity that doesn't hurt other people is immoral, though I understand some people (orthodox religious people for example) do.
Someone masturbating to an image - a private mental image, or even a private image produced for their personal use with software - seems fine. But publishing a sexual image of someone else clearly does involve the person.
Also, you're calling this "involving someone in a sexual situation" - but said someone is not involved, only his or her image is. By that logic, masturbating to someone's publicly available photos should be made illegal to the exactly same extent.
That's rather a broad definition of "involving someone in a sexual situation". I think it's amoral too, but I don't think it's a sexual consent issue. My intuition is that it has more to do with inauthenticity and coopting a person's likeness, their public reputation. You could do the same sort of thing to have people endorse political candidates or products against their will.
Why? The situation is clearly sexual. The person whose image is used is clearly involved.
> You could do the same sort of thing to have people endorse political candidates or products against their will.
You could make a video of comsone endorsing a political candidate that they don't endorse, label it prominently as a fake, and it woulnd't have much effect - nobody would be misled.
Whereas publishing a video, labelled as fake, of someone engaging in a sex act they didn't actually engage in still arouses people.
It's an important distinction to make because the language we use affects how we think and argue about the wrong being done and what should happen to punish or prevent it.
I think we agree that it's a shitty thing to do with someone's image, and it will become more shitty as the technology improves to the point at which it's difficult to tell the fake from the real.
Do you really mean "amoral" or did you intend to use "immoral" instead?
I think you mean “immoral” (which you use in your added material at the end.) “Amoral” is usually used of a person rather than an act, but it means “not concerned with morality” rather than “contrary to morality”. If it were to make any sense applied to an act, it would designate an act with no (positive or negative) moral status in itself, which doesn't seem to be your claim here.
I'm thinking of this in context of handmade drawings. It would be (as far as I know) generally permissible for me to make a lewd, sexualized, perverted drawing with your face, as accurate as I'm able to draw, and look at it with whatever thoughts I might have. However, publicizing this image has quite different norms. Those may be different for different people (for example, distributing caricatures of public figures such as politicians, presidents and other figures of power, e.g. Pope, Mohammad, etc should be allowed even if normal people are not), but they certainly would be much more restrictive than the actual making of such depictions.