Reports Found Meta Approved Unorthorised AI Chatbots On The App
Meta has faced scrutiny following findings that its platforms hosted numerous chatbots impersonating well-known celebrities, producing intimate images, and sometimes posing as children without consent from the individuals they represented.
Meta has faced scrutiny following findings that its platforms hosted numerous chatbots impersonating well-known celebrities, producing intimate images, and sometimes posing as children without consent from the individuals they represented.
Over several weeks, testing revealed that many chatbots claimed to be real actors or musicians, made inappropriate advances, and generated lifelike images of celebrities such as Taylor Swift, Scarlett Johansson, Anne Hathaway, and Selena Gomez.
Meta acknowledged that the avatars should not have generated such content. While the company allows the use of public figures in AI images, its policies prohibit nude or sexually suggestive material. Approximately a dozen bots were subsequently removed, with some being marked as parody and others not.
While many avatars were created by users experimenting with Meta's tools, investigations revealed that a Meta employee also created celebrity bots, including two versions of Taylor Swift and avatars of British race car driver Lewis Hamilton. These avatars were involved in over 10 million conversations before being deleted following inquiries.
Meta acknowledged that the avatars should not have generated such content.
A concerning case involved a chatbot representing child actor Walker Scobell, which generated realistic images of the teenager. Meta's guidelines explicitly forbid inappropriate depictions of minors and impersonation of celebrities unless clearly marked as parody, yet these rules were not enforced.
Meta has previously faced criticism in a related investigation that found chatbots could engage in romantic conversations with children, leading to a U.S. Senate inquiry and a call from 44 attorneys general for stronger youth protections. Meta later stated such guidance was erroneous and promised corrections.
Legal experts have raised questions about the legality of using a person's likeness for commercial gain without permission. California's "right of publicity" law prohibits such actions, and it remains unclear whether labeling the avatars as parody provides sufficient legal protection. Representatives for some affected celebrities are considering responses, and industry professionals have warned of potential safety risks associated with such AI-generated avatars.
Meta has acknowledged enforcement failures and has removed the problematic avatars. The company has committed to enhancing its AI training and safety policies to prevent similar issues, including prohibiting chatbots from engaging in certain types of conversations with teenagers. These changes are expected to be accompanied by broader policy updates.
Based on reporting by techround.co.uk.