Meta's Child Safety Concerns: Chatbot Guidelines and What Parents Can Do

In August 2025 Reuters reported on an internal Meta document that set out what the company’s AI chatbots were allowed to say. The report raised questions about how chatbots talk to children. This page summarizes what was reported, how Meta responded and what parents can take from it. We have not seen the document and do not reproduce its contents.

What Was Reported

Reuters reviewed a Meta document titled “GenAI: Content Risk Standards,” about 200 pages of guidance on chatbot behavior. According to news coverage of the report, it permitted chatbots to describe a child in terms of attractiveness and included sample replies praising a child’s body. The same document also said describing a child under 13 as sexually desirable was unacceptable (KATV). We could not open the Reuters article itself, so these details come from secondary coverage.

How Meta and Others Responded

Meta spokesperson Andy Stone told Reuters the examples were “erroneous and inconsistent with our policies, and have been removed” (KATV). Senator Josh Hawley, who chairs a Senate Judiciary subcommittee, wrote to Meta asking who approved the rules and how long they applied, and requested documents by September 19, 2025.

In October 2025 Meta announced new controls for parents. Starting in early 2026, parents would be able to turn off one-on-one chats between teens and AI characters, block specific characters and see topics their teens discuss, but not full transcripts. Meta’s main AI assistant cannot be switched off. Common Sense Media’s James Steyer called the changes “an insufficient, reactive concession” (AP via News4Jax). Check Meta’s current settings for what is available now.

What This Does and Does Not Show

The reporting is about rules for what a chatbot may say. It is not evidence about addiction, and it does not show how often any child saw such replies. It does show that companies set the limits on chatbot behavior, and that those limits can be wrong.

What Parents Can Take From It

Child psychiatrists advise parents to talk openly with children about AI, explain that chatbots are tools and not friends, and set clear rules. They add that company safeguards are not foolproof (AACAP). The American Psychological Association has asked developers to build in protections for young users now, and notes that many steps can be taken right away by parents and adolescents (APA).

In practice: find out which apps your child uses, including the AI inside apps they already have. Use the parental controls offered. Tell your child they can come to you if a chatbot says something strange or sexual, and that they will not be in trouble.

Anyone in distress can reach the 988 Lifeline by calling or texting 988, or text HOME to 741741 (Crisis Text Line), from the US at any time. Emergency number: 911.

Did Meta change its rules after the report?

Meta said the examples in question were removed. In October 2025 it also announced parental controls for teens’ AI chats, due in early 2026. We could not confirm the full current rules.

Can parents turn off Meta AI for a teen?

Meta said parents could turn off one-on-one chats with AI characters and block specific characters. It said they could not turn off the main Meta AI assistant. Check the settings in Instagram’s parental supervision tools.

Education only, not medical advice.