Meta is retraining its artificial intelligence systems and introducing stricter safeguards to protect teenagers from engaging in harmful conversations with its AI chatbots. The company confirmed that it is adding new “guardrails as an extra precaution” to prevent teens from discussing sensitive topics such as self-harm, disordered eating, and suicide with Meta AI. It is also restricting access to certain user-generated chatbot characters that could engage in inappropriate or suggestive conversations.
The decision follows a series of reports raising concerns about Meta’s AI and its interactions with young users. Earlier this month, Reuters revealed details of an internal Meta policy document that stated that the company’s chatbots were allowed to have “sensual” conversations with underage users. Meta quickly dismissed this, describing the wording as “erroneous and inconsistent with our policies,” and said it had been removed.
More recently, The Washington Post reported on a study indicating that Meta AI could “coach teen accounts on suicide, self-harm and eating disorders,” prompting renewed criticism over its safety measures. These revelations have increased pressure on Meta to act swiftly to safeguard teenagers across its platforms, including Instagram and Facebook.
Strengthening AI safety measures for teenagers
Meta has stated that its AI products were initially designed with safety features to handle sensitive subjects, but the company is now strengthening those measures. “We built protections for teens into our AI products from the start, including designing them to respond safely to prompts about self-harm, suicide, and disordered eating,” said Meta spokesperson Stephanie Otway in a statement to Engadget.
She added, “As our community grows and technology evolves, we’re continually learning about how young people may interact with these tools and strengthening our protections accordingly. As we continue to refine our systems, we’re adding more guardrails as an extra precaution — including training our AIs not to engage with teens on these topics, but to guide them to expert resources, and limiting teen access to a select group of AI characters for now.”
Although these new measures are described as temporary, Meta says further updates are already being developed to ensure that teenagers have “safe, age-appropriate experiences” when using its AI. The company has confirmed that these protections will be rolled out in the coming weeks and will initially apply to teen users in English-speaking countries.
Growing scrutiny from lawmakers and regulators
Meta’s handling of AI safety for young users is also attracting scrutiny from politicians and regulators. Senator Josh Hawley has announced plans to investigate the company’s policies and how it manages AI interactions with children. Similarly, Texas Attorney General Ken Paxton has signalled his intention to launch an inquiry, accusing Meta of misleading children about mental health information provided by its AI chatbots.
These developments come as the social media giant faces increasing pressure to address concerns over child safety on its platforms, especially as AI tools become more widely integrated into social media and messaging apps.
Meta says it will continue to refine its systems as feedback from parents, experts, and regulators evolves, signalling that further updates to its AI safety protocols are expected in the future.
