The great deepfake panic of ’23 in the headlines overlooks just about all the legal issues. The headlines about deepfakes reflect that psychotic level of hysterical comprehension we’ve come to expect from our delightful objectiveless modern media coverage.
The sacred media tradition regarding technical issues remains, however. There’s not a mention of any legal measures to shut down these things. There never is.
That’s kind of strange in this case, because there’s only one way you can source materials to create a deepfake. You need originals, and a way of creating a deepfake “function”, that is, making the deepfake do whatever it is you want it to do.
This isn’t hard. It’s a bit like a 3D model of an object with reference points to the functions, like movements, speech, etc. In short, any deepfake will have all these indicators to prove it is a deepfake. This is also the current version of a lot of very old tech, easy enough to spot.
You also need sound samples to create a deepfake vocabulary, sourced from originals. These originals probably can be traced using voiceprint patterns. Anyway, even if the net collection of sounds is created from scratch, it should have some lineage. People’s voices change a bit over time, and are affected by air, recording equipment, etc. It’s a true forensic exercise, but hardly mysterious.
What’s a crime?
A crime is some form of injury, whether reputational or property or otherwise. Deepfakes go one step better simply by being deepfakes. A person’s image is basically their property, both for privacy and business purposes. Misrepresenting a person can be construed as a breach of privacy and as a possibly damaging misrepresentation in context with the deepfake.
For example, Emma Watson reading Mein Kampf on a deepfake recording could be construed as a few million bucks’ worth of damages. She and many other people could well have probable cause for legal action against publishers.
Watson’s deepfake was published on a forum on that well-known website for geniuses, 4Chan. Whether or not the publisher’s liable for damages is up to a court to decide. The person who made the deepfake, however, almost certainly is in a very difficult legal position. Her voice and image is her property, nobody else’s.
Identifiers
The use of biometrics like images and voiceprints in deepfakes are also probably major issues for breach of privacy and damages. Given that biometrics are used in so many ways as identifiers and do have some legal status as evidence makes their use in deepfakes an issue for law enforcement on multiple levels.
For instance – Can you get deepfake images into security camera or other visual evidence? It’s not that much of a stretch of the imagination, is it? Courts might like to have a say in this not-very-functional range of media evidence, too.
What if deepfake information is planted in physical evidence, in whatever form? Upfront, it’s false evidence; bad enough, you’d think. A bit deeper, and it may involve multiple offenses, trespass, etc.
Crime and deepfakes
The whole issue with deepfakes is that they are fakes. As purported images of actual people, they’re a form of fraud, which is a criminal offense worldwide. You could make movies out of them, or have other uses for them, but so far the main issue is that all uses of deepfakes are either malicious or “threats by media”.
People in the public eye are easy targets for deepfakes. They generate so much visual and other information that creating deepfakes is simple. That applies across the board, and the risks are high.
Even if you allow for the not-very-bright believing Emma Watson reading Mein Kampf is real, there are much more practical options for deepfakes. They can run on phones. They could be mistaken for legitimate communications, like leaving an abusive message.
You could also use celebrity deepfakes to deliver threats, extortion, blackmail, etc. Imagine being blackmailed by Mickey Mouse. The advantage is that the sender doesn’t have to identify themselves, a very common troll tactic.
Again, the ease of finding deepfakes makes them a liability for users. After all – Why would anyone bother to use a deepfake in a real message? The trouble is that the finding of the deepfake happens after the event.
The simple fact is that deepfakes are potentially dangerous. A slightly more complicated fact is that there’s no reason for them to exist at all. Who needs them? You could make a deepfake of yourself and star in a movie, sure. But who else has any right to the image, voice, etc.?
Possession of deepfakes and deepfake software a crime?
In theory, you could make deepfakes a statutory crime. That doesn’t stop much these days, particularly in the US …But it’s a legal status that could be useful in civil and statutory law. A deepfake could have the same status as robbery, which is what it actually is.
What’s legally useful is that it adds a dimension to any other issues. Say you make a deepfake of someone. That someone sues you for breach of privacy. Add to that a criminal offense, and you’re looking at a much stronger finding by the court.
Deepfakes are fixable, but they need a legal status.
_________________________________________________________________
Disclaimer
The opinions expressed in this Op-Ed are those of the author. They do not purport to reflect the opinions or views of the Digital Journal or its members.
