This audio is auto-generated. Please let us know if you have feedback.
YouTube announced expanded access to its likeness detection tools, which use uploaded selfies to alert users to potential misuse of their image in the app.
Originally announced at YouTube’s MadeOn event in September last year, YouTube’s likeness detection process uses face scans, as well as government ID, to reference check uploaded content across the app.
The tool can then alert users to similar visuals within uploaded content, so users can determine whether their image is being used by somebody else.
Users can then submit a removal request if they have a concern. For high-profile users in particular, this offers additional protection.
YouTube initially made likeness detection available to a small number of creators in October. Now, the company is expanding access to a larger pilot group of government officials, journalists and political candidates.
In the age of AI deepfakes, this could be important. Misinformation around the conflict in Iran is already running rife, with misleading content that includes archival footage as well as AI-generated material.
Public figures can also be easily replicated by AI tools now, which makes it important for all platforms to implement some form of protection to ensure that high-profile people are not being misrepresented.
Pew Research data suggests that around half of U.S. adults (53%) now get at least some news from social media, making it an important vector for information. That underlines the need for accuracy in reporting and in being able to stop the spread of false reports whenever possible.
However, some users will have concerns about entering face scans and uploading their ID.
YouTube said incoming users will have to verify their identity before enrolling in likeness detection, though the company clarified in its Tuesday announcement that any data provided during setup is “strictly used for identity verification purposes and to power this safety feature, and is not used to train Google’s generative AI models.”
Maybe that’s enough assurance, and maybe the benefits outweigh the concerns.