Deepfake legal situation: Why the criminal offense is to be introduced

Contents

Films and images of people that appear deceptively real can be created more and more easily with computer technology. The performance of generally available AI models is increasing rapidly, and the results can now compete with real film productions. With easy availability on platforms like X, misuse is also increasing.

Fraud attempts, for example, in which perpetrators use AI tools, are known. These are already punishable by law due to the intent to defraud. The legal situation regarding the misuse of images of private individuals, whether in combination with AI elements or as pure AI products, so-called deepfakes, is less clear.

Debate about a criminal offense gap

The debate about a possible criminal offense gap for pornographic deepfakes has been reignited in Germany after actress Collien Fernandes went public last week with accusations against her ex-husband Christian Ulmen.

Among other things, whether AI providers and platform operators must be held more responsible is being intensely discussed. This is because neither German nor European law currently offers effective protection against such acts – although this does not mean that perpetrators cannot still be prosecuted. However, in practice, this is apparently difficult to enforce.

So far, the various regulations are a colorful puzzle. Above all, the general right of personality has so far been at the center of legal possibilities. However, the claim for removal and subsequent injunction against further distribution is hardly enforceable in practice against the operators and originators if they operate from outside the European legal area. Furthermore, the person demanding removal bears the burden of proof.

Most sexualized deepfakes are not published under real names – obtaining information from the provider about who uploaded them also regularly fails in reality. And other avenues are also ruled out: While the use of copyrighted works may be a legal violation or the distortion of an original image, claims on this basis are hardly enforceable in reality.

Criminal offense gaps in deepfakes

For this reason, the state of Bavaria had submitted a regulation proposal for a new paragraph 201b of the Criminal Code in 2024 via the Bundesrat: “Violation of personality rights through digital forgery.” Systematically, it would follow the regulatory content of §201a, which penalizes “violation of the most personal sphere of life and personality rights through image recordings” with imprisonment or a fine. This is precisely where the Federal Ministry of Justice (BMJV) likely intends to start with its expected proposal.

The Bavarian proposal from 2024 concerned making accessible “media content produced or altered by computer technology” if it gave the appearance of “realistic image or sound recordings.” This definition was very broad and is unlikely to be repeated in the upcoming formulation by the BMJV.

One problem would remain: the creation itself would still not be criminally punishable. This is because criminal liability would only arise from distribution – for example, sending via messenger or uploading to platforms. What someone does locally on their end device with photos or video material and saves it there would not be covered.

Creation is to become punishable

Federal Justice Minister Stefanie Hubig (SPD) now wants to change this. The criminal offense is to be committed and thus prosecutable already with the creation of sexualized deepfakes. Thus, the acquisition and possession of relevant material would also be punishable, not just active distribution.

If a new §201b StGB were to follow the systematic approach of §201a, it would also be expected here that it would generally be a so-called “Antragsdelikt” (offense requiring a formal complaint): Affected parties – in most cases women – would then have to file a request for prosecution. Only in exceptional cases would prosecution be possible due to public interest. Nevertheless, Hubig is counting on the deterrent effect of criminal liability.

However, with a criminal regulation for the creation and distribution of deepfakes, a balancing act will always have to be performed in the future as to whether a deepfake is not permissible. While a legal form of use is rarely likely in the case of sexualized depictions, this is much more possible with – sometimes tasteless – satirical contributions. And criminal law must also draw fine distinctions with media, art, and science freedom to largely exclude unintended collateral damage.

Meanwhile, providers themselves are left out of domestic legislation: Liability regulations for providers are conclusively regulated at the European level – especially the Digital Services Act (DSA) plays a role here. However, an amendment to the Criminal Code could also oblige providers to intervene more quickly: providers must examine reported offenses.

In addition, established procedures between providers and law enforcement agencies for securing evidence could be used, which could increase the pressure on perpetrators in the future. However, much will also depend on the willingness of the platforms to cooperate. Whether they will then also suppress the re-publication of already objected-to content is hardly reliably predictable.

This is just one reason why politicians and supervisory authorities in the EU are calling for: The operators of generative AI models should suppress the creation of sexualized content from the outset if their models could be used for this purpose.

(akn)