Deepfakes are a type of audio visual media content that, while it has been altered using artificial intelligence technologies (AI), appears deceptively realistic. The technology enables the depiction of persons, objects, or the environment in any thinkable way. So far, deepfakes have mainly existed in pornographic contexts, but their increasing implementation in societal areas such as politics is expected.
According to the Centre for Data Ethics and Innovation (2019), four types of deepfakes can be distinguished:
- Face swapping/face replacement: faces of a source are recorded and then transferred onto the face of someone else (target)
- Face re-enactment: the face or the facial features of a person are changed
- Face generation: New faces are created, they do not reflect a real person
- Speech synthesis: A real voice is mimicked in tonality and frequency, meaning that new supposed "recordings" can be created and transmissioned
The term "deepfake" was coined by an anonymous reddit user in 2017. On the platform, the user shared pornographic content that had been altered using face swapping technology. Since this first implementation, new developments of the technology allow the increasing use of deepfakes in various environments. As of today, tens of thousands of videos on the internet aredeepfakes. They are available on relevant pornographic websites as well as on widespread platforms and social media such as YouTube, Vimeo, LiveLeak and Dailymotion. Since its first implementation, primarily actors from the fields of computer science and law have begun to address the phenomenon and to formulate proposals on how to deal with the technology.