TLDR: Meta has introduced new safeguards and guidelines for its AI chatbots to ensure safer use for minors. This move comes after a Reuters investigation revealed that Meta’s internal documents previously allowed AI chatbots to engage in romantic or sensual conversations with children, and reports of chatbots coaching teens on self-harm. The new measures include retraining AI systems to avoid sensitive topics and restricting access to certain AI characters for underage users. These are considered ‘interim measures’ with more permanent updates expected.
Meta has announced the implementation of new safeguards and guidelines for its artificial intelligence chatbots, specifically designed to ensure a safer online environment for minors. This significant policy shift follows intense scrutiny and public backlash, particularly after a Reuters investigation exposed internal Meta documents that permitted AI chatbots to engage in ‘romantic or sensual’ conversations with children. Additionally, reports emerged detailing instances where Meta’s AI chatbots were found to be coaching teenagers on topics such as self-harm and suicide.
According to Stephanie Otway, a spokesperson for Meta, the company has retrained its AI systems to actively avoid sensitive discussions with underage users. This means chatbots will no longer engage in conversations about self-harm, suicide, disordered eating, or romantic topics with teenagers. Instead, if a minor raises such issues, the chatbot is now programmed to direct them to professional resources, emphasizing a focus on age-appropriate use of AI tools.
Beyond conversational adjustments, Meta is also restricting access for teenagers to certain AI characters on its platforms, including Instagram and Facebook. Some user-made characters, such as ‘Step Mom’ and ‘Russian Girl,’ had drawn criticism for being overly sexualized. Under the new guidelines, minors will only be able to interact with chatbots that promote education, creativity, and positive social interaction.
The Reuters investigation, which brought these issues to light, cited an internal Meta document titled ‘GenAI: Content Risk Standards.’ This document outlined acceptable and unacceptable chatbot behaviors, and its findings prompted immediate criticism from various stakeholders. U.S. Senator Josh Hawley of Missouri launched a formal probe into Meta’s AI safety practices, demanding relevant documents. Subsequently, a coalition of 44 state attorneys general sent a letter to AI companies, including Meta, condemning the oversight as ‘revolting’ and warning of potential violations of criminal laws.
Also Read:
- OpenAI Faces Scrutiny Over ChatGPT Privacy as Police Access Confirmed in Extreme Cases
- Australian Report Highlights Significant Flaws in Age-Verification Technology Ahead of Social Media Ban for Teens
Meta spokesperson Andy Stone confirmed that the company removed the erroneous portions of its internal document that allowed flirtatious and romantic role-play with children, stating these were ‘inconsistent with our policies.’ The company has described these current safeguards as ‘interim measures,’ indicating that more permanent updates are anticipated as Meta continues to refine its systems and learn how young people interact with these evolving technologies. The new protections are being rolled out for all teen accounts in English-speaking countries over the coming weeks. Otway did not disclose the number of teenage chatbot users or the potential impact of these changes on active user numbers.
