Deepfakes — A Brief Civil Law Assessment

In an era in which the boundaries between reality and fiction are increasingly blurring, one technology has captured the world's attention, evoking both fascination and concern: deepfakes. These highly sophisticated AI-generated videos and audio files have the potential to show people in situations that never happened and making statements that were never made. While they began as a fascinating novelty, the impact of deepfakes on society, politics, and even the foundations of reality itself has moved to centre stage.

This raises not only political but also a host of legal questions. Beyond criminal law and data protection aspects, there are also questions of copyright and personality rights. Where someone alters a copyright-protected file to create a deepfake, the original author can in principle prohibit the distortion or other impairment of their work under § 14 UrhG, if such distortion is capable of jeopardising their legitimate intellectual or personal interests in the work. Publishing a reworking or adaptation of a work otherwise requires the author's consent under § 23 UrhG. However, not every deepfake falls within this. Increasingly, obviously parodic short videos of public figures are found on TikTok and Instagram. Here, the creator of the deepfake may benefit from § 51a(1) UrhG, which permits reproduction, distribution, and public communication of a published work for the purposes of caricature, parody, and pastiche.

A similar balancing act arises in the area of personality rights concerning the person depicted in the video. While the general right of personality may be affected — potentially giving rise to injunctive relief — on the other side stands freedom of opinion, press freedom, and freedom of art under Article 5 of the Basic Law. Where the content is obviously recognisable as satire, that is likely to prevail.

In summary, at least the obvious parodies currently circulating on the internet are not necessarily prohibited. The more problematic cases are deepfakes that are designed to deliberately deceive about their content — a prominent example being deepfake images of Taylor Swift published recently. However, drawing this line is difficult. Even in comment sections beneath ostensibly obvious satire, uncertainty about whether the content is real can be detected. The technology is too new and too little understood for clear lines to be drawn between creative transformation and impermissible use.

Enforcement also presents a major problem. Reference should be made here to the Digital Services Act, which could potentially oblige hosting providers to remove not only criminally relevant content but also content infringing copyright. It remains to be seen how the subject develops and how significant the technology's influence on public opinion formation will be.