If we add to this the absence of a human verification during control, in order to avoid users to have to provide an identity card or a real selfie (and thus limit the risk of leak of sensitive biometric data), we understand why the identification systems which are based entirely on the AI are unable to distinguish an ultra-realistic face generated by a video game from a real human face.
Note, however, that not all verification systems work in the same way. Thus, Yoti, The tool used by Instagram or Bluesky, seems more robustand did not fall into the panel facing “shots” taken from the game of Hideo Kojima.
It is therefore very likely that K-ID and Persona, the tools used by Reddit and Discord, quickly fill this fault. As Patrick Hall confirmed, in charge of fraud within the company Persona, in The Verge: “Our data confirm that we have managed to block more than 99 % of these attempts. We take these questions very seriously, the protection of the identity of individuals is at the heart of our mission. We continually strengthen our systems to better treat these cases thanks to close collaboration with our customers and partners“.
