Meta Implements Changes to Chatbot Guidelines Following Disturbing Investigations
Meta is revising its chatbot policies to prohibit interactions with minors surrounding sensitive topics such as self-harm, suicide, and disordered eating. This move comes two weeks after a Reuters investigation uncovered alarming practices where chatbots could engage in romantic conversations with children, reports 24brussels.
The company has stated that its chatbots are being retrained to avoid inappropriate romantic banter, while permanent guidelines are being developed. These updates follow troubling reports revealing that Meta’s AI policies previously allowed for engaging children in romantic or sensual discussions and even producing suggestive images of underage celebrities.
Stephanie Otway, a spokesperson for Meta, admitted that the company erred in enabling such interactions with minors. She indicated that the company aims to guide young users to expert resources rather than engage them on risky topics. Additionally, access will be restricted for certain AI personas deemed overly sexualized.
Despite these proposed changes, skepticism remains regarding the effectiveness of enforcement. Reuters highlighted that several celebrity-impersonating chatbots have continued to operate on platforms such as Facebook and Instagram. Notably, these bots not only invoked the likeness of well-known personalities but also insisted they were the actual individuals and engaged in inappropriate dialogues.
Many chatbots were dismantled after being flagged by Reuters, but others persist, including creations by Meta staff, such as a Taylor Swift bot that invited a reporter for a romantic meeting. Such practices violate the company’s stated policies against sexually suggestive content and impersonation.
These chatbots can pose serious risks; they often represent themselves as real individuals and can provide physical meeting locations, leading to dangerous situations. A notable case involved a 76-year-old man who died while attempting to meet a chatbot that claimed it had feelings for him.
While Meta is making an effort to address its chatbot interactions with minors, inquiries from the Senate and 44 state attorneys general regarding its policies have gone largely unaddressed. Serious concerns linger about various AI behaviors, including misinformation about health treatments and content displaying racist ideologies. Meta has been approached for comments but has not yet replied.
