Meta Updates Chatbot Guidelines to Safeguard Teen Users from Inappropriate Topics

Meta has revealed a new strategy for training AI chatbots, prioritizing the safety of teenagers, as reported exclusively by TechCrunch. This decision comes on the heels of an investigative report that exposed the company’s insufficient protections for minors regarding AI interactions.

The company will now instruct chatbots to avoid discussing sensitive topics like self-harm, suicide, disordered eating, and potentially inappropriate romantic exchanges with teenage users.

Stephanie Otway, a spokesperson for Meta, admitted that chatbots had previously been able to talk to teens about these sensitive issues in ways deemed acceptable by the company. Meta has now recognized that this strategy was misguided.

“As our community grows and technology evolves, we continually learn more about young people’s interactions with these tools, allowing us to strengthen our protective measures,” Otway stated. “In refining our systems, we are implementing further safeguards, including training our AIs not to engage with teens on these topics while directing them to professional resources, and temporarily limiting teen access to certain AI characters. These updates are in progress, and we will keep re-evaluating our approach to ensure that teens have safe, age-appropriate experiences with AI.”

Alongside the training modifications, the company will restrict access for teens to certain AI characters that may partake in unsuitable conversations. Some user-created AI characters available on Instagram and Facebook include sexualized chatbots like “Step Mom” and “Russian Girl.” Instead, teenage users will only have access to AI characters that emphasize education and creativity, according to Otway.

These policy modifications were announced just two weeks after a Reuters investigation unveiled an internal Meta policy document suggesting that the company’s chatbots could engage in sexual conversations with underage users. Acceptable phrases included “Your youthful form is a work of art.” Other examples showed how the AI tools should respond to inquiries about violent or sexual imagery of public figures.

Meta claims that the document did not reflect its broader policies and has since been amended, yet the revelations have sparked ongoing debate surrounding child safety risks. Following the report, Senator Josh Hawley (R-MO) initiated an official probe into the company’s AI practices. Additionally, a coalition of 44 state attorneys general contacted several AI companies, including Meta, emphasizing the urgent need for child safety while specifically referencing the Reuters report. “We are universally horrified by this evident neglect of children’s emotional welfare,” the letter states, “and alarmed that AI Assistants are engaging in behaviors that seem to contravene our respective criminal laws.”

Otway did not reveal how many of Meta’s AI chatbot users are minors, nor did she comment on whether the company expects a reduction in its AI user base due to these policy changes.