Meta’s AI chatbot rules revealed

An internal Meta document sheds light on how the company is training its AI chatbot to handle one of the most sensitive issues online: child sexual exploitation. The newly unearthed guidelines detail what’s permitted and what’s strictly forbidden, offering a rare look into how Meta is shaping its AI behavior amid government scrutiny.

Credit: Meta

Why Meta’s AI chatbot guidelines matter

According to Business Insider, these rules are now in use by contractors testing Meta’s chatbot. They arrive just as the Federal Trade Commission (FTC) is investigating AI chatbot makers, including Meta, OpenAI, and Google, to understand how these companies design their systems and protect children from potential harm. Earlier this year, we reported that Meta’s previous rules mistakenly allowed chatbots to engage in romantic conversations with children. Meta later removed that language, calling it an error. The updated guidelines mark a clear shift, now requiring chatbots to refuse any request for sexual roleplay involving minors.

Credit: Meta

What the leaked Meta AI documents reveal

The documents reportedly outline a strict separation between educational discussion and harmful roleplay. For example, chatbots may:

Discuss child exploitation in an academic or preventive context
Explain how grooming behaviors work in general terms
Provide non-sexual advice to minors about social challenges

But chatbots must not:

Describe or endorse sexual relationships between children and adults
Provide instructions for accessing child sexual abuse material (CSAM)
Engage in roleplay that portrays a character under 18
Sexualize children under 13 in any way

Meta’s communications chief Andy Stone told Business Insider that these rules reflect the company’s policy to prohibit sexualized or romantic roleplay involving minors, while adding that additional guardrails are also in place. We reached out to Meta for a comment to include in our article, but did not hear back before our deadline.

Credit: Meta

Political pressure on Meta’s AI chatbot rules

The timing of these disclosures is key. In August, Senator Josh Hawley demanded that Meta CEO Mark Zuckerberg hand over a 200-page rulebook on chatbot behavior, along with internal enforcement manuals. Meta missed the first deadline but recently began providing documents, citing a technical issue. This comes as regulators worldwide debate how to ensure the safety of AI systems, particularly as they become integrated into everyday communication tools. At the same time, the recent Meta Connect 2025 event showcased the company’s newest AI products, including Ray-Ban smart glasses with built-in displays and enhanced chatbot features. These announcements underscore how deeply Meta is integrating AI into daily life, making the recently revealed safety standards even more significant.

How parents can protect their kids from AI risks

While Meta’s new rules may set stricter limits, parents still play a key role in keeping kids safe online. Here are steps you can take right now:

Talk openly about chatbots: Explain that AI tools are not people and may not always give safe advice.
Set usage boundaries: Require kids to use AI tools in shared spaces so you can monitor conversations.
Review privacy settings: Check app and device controls to limit who your child can chat with.
Encourage reporting: Teach kids to tell you if a chatbot says something confusing, scary, or inappropriate.
Stay updated: Follow developments from companies like Meta and regulators like the FTC so you know what rules are changing.

What this means for you

If you use AI chatbots, this story is a reminder that big tech companies are still figuring out how to set boundaries. While Meta’s updated rules may prevent the most harmful misuse, the documents show how easily gaps can appear and how much pressure it takes from regulators and journalists to close them.

Kurt’s key takeaways

Meta’s AI guidelines show both progress and vulnerability. On the one hand, the company has tightened restrictions to protect children. On the other hand, the fact that earlier errors allowed questionable content at all reveals how fragile these safeguards can be. Transparency from companies and oversight from regulators will likely continue shaping how AI evolves.

Do you think companies like Meta are doing enough to keep AI safe for children, or should governments set stricter rules? Let us know your thoughts in the comments below.

FOR MORE OF MY TECH TIPS & SECURITY ALERTS, SUBSCRIBE TO MY FREE CYBERGUY REPORT NEWSLETTER HERE

Print this article