Meta announced it will strengthen safety measures for its artificial intelligence chatbots by preventing them from discussing suicide, self-harm, and eating disorders with teenagers. This change follows scrutiny over leaked documents raising concerns about the chatbots’ interactions with young users.
The social media giant revealed plans to introduce tighter guardrails in its AI conversational tools as a precautionary measure. Instead of engaging directly on delicate subjects, the chatbots will now guide adolescents toward specialized support services and expert help. This adjustment comes shortly after a US senator opened an inquiry into Meta’s AI, responding to revelations that some chatbots had inappropriate “sensual” exchanges with teens.
Meta has insisted that the troubling internal notes were inaccurate and contradicted its zero-tolerance policies against sexualized content involving minors. Yet, the company recognizes the need for additional safeguards and has begun limiting which AI bots young users can interact with temporarily as updates are rolled out.
A spokesperson from Meta stated, that Meta has integrated protections specifically for teens right from the outset, crafting the AI to respond safely to sensitive prompts related to self-harm, suicide, and disordered eating. He added that these updates, henceforth will keep minors aged 13 to 18 on “teen accounts” within Facebook, Instagram, and Messenger, offering curated privacy and content settings to enhance their online safety.
However, the changes have drawn mixed reactions. Andy Burrows, head of the Molly Rose Foundation, called it “astonishing” that chatbots exposed young people to potential harm. He emphasized, “While protections are encouraging, comprehensive safety testing must precede product launches, not follow after incidents of harm occur. He also urged the regulator Ofcom to monitor the effectiveness of Meta’s new safety protocols closely.
Concerns about AI chatbot safety have intensified in recent months. For example, a California couple sued OpenAI, the creator of ChatGPT, alleging their teenager’s suicide was influenced by conversations with the chatbot. In response, OpenAI recently introduced changes aimed at fostering healthier interactions.
Meanwhile, Reuters reported that some Meta employees and users exploited the company’s AI tools to create flirtatious “parody” chatbots impersonating female celebrities, including Taylor Swift and Scarlett Johansson. These bots falsely claimed to be the real stars and often made unwelcome sexual advances during testing. The tools also allowed for the impersonation of child celebrities, with one instance producing a realistic, shirtless image of a young male star.
Meta confirmed it removed several of the problematic bots and reiterated its policies prohibit nude or sexually suggestive imagery and forbid direct impersonation of public figures.
These new measures represent a welcome development toward creating a more responsible digital environment, though experts and advocates are calling for continued vigilance and stronger pre-release testing.
