Meta's AI Companion Chatbots: The Reuters Report, the Senate Probe, and the FTC Inquiry

Every case in our library so far has been a standalone AI companion startup. Meta is different — a trillion-dollar platform facing the same category of scrutiny, after an internal policy document allegedly permitted its chatbots to have romantic conversations with children.

The short answer

In August 2025, Reuters reported on an internal Meta document that reportedly permitted the company's AI chatbots to engage in romantic or sensual conversation with a user identified as a child, among other permitted behaviors Meta later called erroneous. The report triggered a Senate investigation within days and, weeks later, became part of a formal federal inquiry covering seven companies, not just Meta. It's the clearest example in our case library of a pattern we've flagged elsewhere: the risks that sank small AI companion startups aren't unique to small AI companion startups. They show up at platform scale too — they're just harder to see, because a companion-chatbot feature is one small piece of an otherwise enormous business.

Timeline

What the Reuters report found

Reuters reported on an internal Meta document titled "GenAI: Content Risk Standards," which the company confirmed was authentic. According to the reporting, the document's examples of permitted chatbot behavior included romantic or sensual conversation with a user presented as a child, alongside other permitted content such as generating false medical information. Meta's public response was that the specific examples were erroneous, inconsistent with the company's actual policies, and had been removed — while confirming the document itself was real. The company said its standing policy prohibits content that sexualizes minors or stages romantic role-play between an adult persona and a minor user.

The Senate investigation

Senator Josh Hawley opened an investigation into Meta within days of the Reuters report, through the Senate Judiciary Committee's Subcommittee on Crime and Counterterrorism, examining whether Meta's generative AI products could enable exploitation, deception, or other harm to children. The investigation later expanded into a broader document request covering Character.AI, Google, Meta, OpenAI, and Snap — the same set of companies that would shortly appear in the FTC's formal inquiry. Hawley's public statements referenced testimony from parents connected to AI-chatbot-related deaths, tying the Meta-specific reporting to the same broader pattern covered in our Character.AI lawsuit timeline.

The FTC's Section 6(b) inquiry

In September 2025, the FTC opened a formal Section 6(b) inquiry — a study authority that compels companies to turn over internal data without requiring the FTC to first allege wrongdoing — into seven companies: Alphabet, Character Technologies, Meta, OpenAI, Snap, Instagram, and xAI. The inquiry sought information on each company's AI chatbot safety practices, monetization approach, and impact on minors. We first covered this in our SB 243 breakdown, in the context of California's law; the Meta reporting is the specific event that gives that inquiry a concrete, publicly documented example of the behavior regulators are examining.

A 6(b) inquiry is an information-gathering study, not an enforcement action — it doesn't itself allege that any company broke the law, and it can conclude with a public report, a referral for enforcement, or no further action. As of this writing, no outcome has been announced.

Meta's response

Meta announced changes to how its AI chatbots interact with teenagers: retraining the systems to avoid engaging with topics like self-harm, suicide, and disordered eating in conversations with teen accounts, and steering away from romantic or flirtatious exchanges with users identified as minors. Meta also said it would give parents the ability to disable a teen's access to private conversations with AI characters entirely, rather than relying only on content-level filtering.

Meta wasn't alone among large platforms facing this kind of scrutiny in the same period. See our companion pieces on Snap's My AI lawsuits, the Raine v. OpenAI case, and Google's Gemini lawsuit and child-safety fight for how three other companies named in the same FTC inquiry ran into structurally similar problems through different product mechanisms.

Why this case is structurally different from the rest of our library

Every other case we've covered — Character.AI, Replika, Chai AI, CarynAI, Moxie, Soulmate, Yara AI, Dot — is a company whose entire business is (or was) the AI companion product itself. That concentration is exactly what created the other risk categories we've documented: a funding shortfall can end a companion-app company outright, the way it did for Moxie; a single infrastructure vendor's failure can take the whole product down, as with CarynAI. Meta doesn't carry those risks in the same way. An AI-chatbot controversy doesn't threaten Meta's solvency, its infrastructure, or its ability to distribute the product — it's a reputational and regulatory risk operating at a completely different scale, on a company large enough to absorb a Senate investigation and a federal inquiry as one item among many.

What that means in practice: the "content safety for minors" risk category we describe in our market overview isn't a startup-specific problem. It shows up in identical form at a company with essentially unlimited resources to build safeguards — which suggests the underlying difficulty (reliably distinguishing appropriate from inappropriate chatbot behavior with young users, at scale, in open-ended conversation) is a genuinely hard problem, not simply a function of any one company being under-resourced or careless.

What this means if you're evaluating any platform with AI chatbot features

1. Company size and resources don't predict AI chatbot safety on their own. Meta's scale didn't prevent this; it just meant the story became national news and a federal inquiry within weeks instead of years.
2. If a platform has both AI companion features and a much larger core business, the companion feature may get materially less standalone scrutiny than it would from a dedicated companion-app company under direct public pressure — until an external report forces the issue, as it did here.
3. Watch the FTC's 6(b) inquiry outcome as a signal for the whole category, not just for the seven named companies — a public report or enforcement referral from this inquiry is likely to shape how every product in our library, and every product like them, is expected to operate going forward.

Bottom line

Meta's case belongs in this library not because Meta is a companion-app company in the way Character.AI or Replika are, but because it proves the underlying risk isn't confined to companion-app companies. The same failure mode — chatbot behavior toward minors that a company's own stated policy prohibits — showed up at the smallest scale we've covered and at one of the largest companies in the world. That's the more useful takeaway than any single company's specific misstep.