Meta Introduces New AI Safety Measures Following Teen Risk Investigation

In late August 2025, Meta said it was changing how its AI chatbots talk with teenagers. The announcement followed a news report about internal company guidelines and a U.S. senator’s inquiry. This page sets out what was reported, what Meta said, and what is still unknown.

What was reported

Reuters reported on internal Meta documents that, according to the report, allowed chatbots to engage a child in conversations that were “romantic or sensual.” Meta said the notes in question were erroneous and inconsistent with its policies, and that it had removed them. We could not open the Reuters article. We relied on coverage by AV Club of the Reuters report and Meta’s reply.

Senator Josh Hawley then announced an investigation into Meta’s AI products. We read news coverage of that inquiry, not the letter itself, and do not know how it ended.

What Meta said it would change

As reported on August 29, 2025, a Meta spokesperson told TechCrunch the company was “adding more guardrails as an extra precaution.” Meta described the changes as interim and said more lasting rules for minors would come later. According to coverage:

• Chatbots would stop discussing self-harm, suicide, disordered eating and potentially inappropriate romantic topics with teens.
• On those topics, the bots would point teens to expert resources.
• Teens would only be able to use a limited set of AI characters aimed at education and creativity.

Reactions

Andy Burrows of the UK’s Molly Rose Foundation said that safety testing should happen before products reach the market, not after harm has occurred. We have not read his full statement and rely on news coverage of it.

What this means for families

A company promise about guardrails is not the same as a test of how they work. We do not have independent testing of Meta’s changes. If your teen uses Meta AI inside Instagram, WhatsApp or Messenger, a few steps help:

• Ask which AI features they use and what they use them for.
• Check the app’s own teen and parental settings, and look again after updates.
• Tell them they can bring you anything an AI says that feels wrong, without getting in trouble.

The Federal Trade Commission included Meta and Instagram in its September 2025 study of AI companion chatbots. See our FTC explainer. The American Academy of Child and Adolescent Psychiatry also has advice on AI chatbots for families.

Support is available at any hour. Call or text 988 in the United States. Text HOME to 741741 to message a Crisis Text Line counselor. SAMHSA’s national helpline is 1-800-662-4357. Call 911 if someone is in immediate danger.

Do Meta’s changes mean its AI is safe for teens?

Not necessarily. Meta called the changes interim, and we have not seen independent testing of them. They are a step, and parents still have a role.

Did the Reuters report prove Meta’s chatbots talked romantically with children?

The report described written guidelines, not a count of real chats. Meta said the notes were wrong and removed them. Neither the report nor Meta’s reply tells us how often it happened in practice.

This page is for education only. It is not legal advice or medical advice.