The study “Deepfakes and Manipulated Realities“ makes it clear right from the beginning of its conclusion that “deepfake“ technologies are likely to have a firm place in everyday life in the future and speaks of a future “altered media reality“ with regard to deceptively real fakes in audio or video form. However, deepfakes or “synthetic media“ are still relatively rare at the moment, which is why the study first sheds light on the current state of deepfake technologies and examines the level of knowledge about them among the Swiss population.

It is becoming increasingly simple to create deepfake content

The technical production of deepfakes has made considerable progress in recent years and is likely to develop rapidly in the coming years. A wide range of deepfake content can already be produced today using AI-based text, image or audio generators – provided that sufficient data material (e.g. speech samples or portrait photos) and sufficient processing capacity are available. It can then be used to imitate a person's way of speaking, facial expressions and even body movements. High-quality deepfake audio content can currently be produced with comparatively little effort. The creation of high-quality deepfake videos currently still involves a great deal of effort, but rapid progress in the development of AI video generators will make it easy to create credible deepfake videos in the future.

Using a representative survey of over 1,300 people, the study investigated how the Swiss population perceives deepfakes, what experiences they have had with them and how they see the opportunities and risks of deepfake technologies. The study results showed that “deepfake“ was still an unfamiliar term for many of those surveyed – only around half had heard of it or seen deepfakes themselves. Only a very small minority had already had experience of creating (2%) and disseminating (3%) deepfakes. An experiment conducted as part of the study showed that the respondents were barely able to distinguish deepfakes from real videos, especially if they were of good or very good quality. The respondents also strongly associated deepfakes with risks and less with opportunities. However, the terms chosen also determined how the debate developed: When “synthetic media“ was mentioned, respondents' assessments were more positive than when “deepfakes“ were referred to.

Primarily, deepfakes are seen as a risk

Furthermore, the study also examined the current and future role of deepfakes in judiciary, journalism, politics and business, this involved expert interviews and surveys being conducted with representatives of the respective areas or literature research. In journalism and politics, deepfakes are primarily seen as a risk and are often categorized as disinformation. The journalists interviewed particularly emphasized the importance of sensitizing the audience to deepfakes and the need for passing on knowledge in order to verify content and its authenticity. In a separate survey, Swiss parliamentarians and members of the federal administration expressed concern that deepfakes could have a negative impact on the officials themselves, democracy and political institutions, and that there are currently very few real protective measures in place.

However, the analyses of the legal and economic sectors show that deepfakes can represent more than just risk and disinformation: While some applications of deepfake technologies are clearly illegal, in other cases deepfakes are protected by freedom of expression, freedom of information and artistic freedom or copyright law. When looking at the use of deepfake technologies in business in particular, the study makes it clear that they also have enormous creative and economic potential. Nevertheless, the interviewees from the business world certainly recognize the danger posed by deepfakes and point to possible uses in the area of white-collar crime such as identity fraud and identity theft, damage to the reputation of companies, deception of employees or uses in the context of economic and industrial espionage.

The study makes four key recommendations

Dr. Murat Karaboga, author of the deepfake study at Fraunhofer ISI, derives the following recommendations for action from the study: “Our study shows that a combination of different measures is required to address the challenges caused or aggravated by deepfakes. The state, the general public, organizations and the media need to all work hand in hand. An initial recommendation following our study is therefore that government efforts to regulate platforms, which also include the deletion or blocking of reported deepfakes and a reporting system for illegal content, including corresponding transparency requirements, should be promoted as they are essential for a sustainable approach to deepfake technologies in Switzerland. Secondly, through the provision of educational programs, individuals should be empowered to take responsibility, with a particular emphasis being put on victim support counseling centers and promoting and teaching media literacy. Thirdly, organizations in all social and economic sectors should prepare themselves for deepfakes turning up everywhere and, for example, carry out internal risk assessments and develop both preventive and reactive measures. A fourth recommendation is aimed at journalism, since by upholding journalistic standards it can help to better recognize deepfakes and immediately inform the public about fake content. Finally, in view of the enormous speed of the technological development, it is also important that politics, business and society should become aware of the speed with which change is taking place and be prepared to identify new developments and react with the appropriate action and additional measures.“