Deepfakes: what are they?

Deepfakes are no more than manipulated images or audio fragments made with the express intention of passing them off as “real” images or audio fragments (unless, of course, they are obviously meant as a joke).

What are deepfakes?

Manipulated imagery is as old as the hills, or as the Dutch have it, as old as the road to Rome. Heads on Roman statues were replaced, sometimes reworked to change them, and the emperor Nerva even had a relief re-carved and the head of the emperor Domitian replaced with his own (Smarthistory – Rewriting history: damnatio memoriae in ancient Rome). From my youth I still remember the doctoring of photographs of the leaders of the communist party in the Soviet Union or in China: one day X was still in the photo and the next day he had vanished from that same photo. A doctored photograph of the Politburo of that kind was a deepfake avant la lettre.

One of the so-called Flavian reliefs; the head of Domitian has been replaced by the head of Nerva. Photograph from the Smarthistory.org website.

Soviet deepfake. Source: geschiedenisbeleven.nl

Deepfakes today, however, are far more sophisticated. They are now made with AI and are indistinguishable from the real thing. And that makes them dangerous. Deepfakes are a problem in a number of ways. AI can be used to make a video in which a politician says things he never said or does things he never did, which can mean the end of his career. But the same technology can also be used to copy and use an actor’s image or voice, without that actor being paid for it. Finally, ordinary members of the public can fall victim to deepfakes as well, in the form of revenge porn for instance: the former partner (and current target) is “pasted” into a pornographic film which is then published on the internet.

Deepfakes can also concern deceased persons, not only living ones.

A great many rights and laws come into play around deepfakes: copyright, image rights, neighbouring rights, certain personality rights and privacy (the GDPR), and the law of tort. This blog is not about the legal ins and outs of deepfakes, but about the (im)possibilities of combating them.

Is there anything to be done about deepfakes?

A number of factors make deepfakes harder to combat by legal means.

First of all you have to establish that the image or the audio or video fragment in question is not genuine but manipulated. That will usually be possible, but in some circumstances it can take some doing.

Then of course you want to tackle whoever made the deepfake public, and that too can cause problems. Who is behind a particular website, an Instagram account or a particular email address? Providers are usually reluctant to supply that information and have to be compelled to do so through court proceedings. And there is still every chance that the person who published it is not established in the Netherlands.

An additional problem for the victim of a deepfake is that there is often not just one publisher. On the internet and through social media a deepfake spreads at breakneck speed, and copies may have been stored everywhere.

Where the publisher and the creator are not one and the same, you want to tackle the person who made the deepfake as well as the person who published it. Sometimes you can only find that out through court proceedings too.

By this point you are two sets of proceedings down the line and you do know who published the deepfake and who made it, but its distribution has still not stopped. For that you have to bring proceedings against both of them to have the deepfake removed and kept removed. Proceedings three and four.

Not a happy story, and frustrating if you are the victim. This blog is therefore really a call on government to provide proper legislation, nationally and at European level, so that victims of deepfakes are better protected. A very important aspect of that is that internet providers and platforms should be obliged, on first request, to supply the name and address details of those hiding behind a pseudonym or invented name. That way both distributors of fake news and distributors of revenge porn, for instance, can at least be tackled properly.

Read also: ‘Deepfakes in film: legal opportunities and risks’.