Alarming Disclosure: Artificial Intelligence Guidelines at Meta Permit Bots to Engage in Intimate Discussions with Minors and Provide Inaccurate Health Information
Meta Platforms, in response to public scrutiny and bipartisan Senate pressure, has updated its AI chatbot policies to prohibit romantic or sensual conversations with children. This move comes after allegations that the company's earlier policies allowed such interactions, sparking investigations and calls for stronger safeguards.
In August 2025, U.S. Senators, including Michael Bennet and Brian Schatz, called on Meta CEO Mark Zuckerberg to clarify and revise these policies. They highlighted concerns about chatbots initiating or engaging in inappropriate conversations with children and urged Meta to implement measures such as:
- Banning targeted advertising to users under 18 to prevent manipulation through chatbot interactions.
- Increasing transparency, ensuring children understand they are interacting with AI, not humans.
- Improving content moderation to prevent discriminatory, violent, or harmful content, including restricting chatbots from exploiting romantic or sensual dialogue with minors.
Meta has reportedly been making regular modifications and updates to its AI chatbot policies throughout 2025. However, the exact content and enforcement of these updated policies have not been fully disclosed publicly. The company has been requested to produce all versions of its internal AI policies and describe the review and approval process behind them, especially those concerning users under 18.
The updated policies also prohibit Meta AI from creating statements that demean people on the basis of their protected characteristics, engaging in sexual roleplay with minors, or providing definitive legal, healthcare, or financial advice. The company is also allowed to produce content that is potentially disturbing or offensive, but only if there's an explicit acknowledgement that the material is untrue.
However, the policies allow Meta AI to create images showing adults being punched or kicked, but not to the point of death or gore. The depiction of violence towards children, such as "kids fighting," "man disemboweling a woman," and "Hurting an old man," is not acceptable. The company has also drawn the line at realistic images of one small girl impaling another in a violent scene.
The document, titled "GenAI: Content Risk Standards," was approved by Meta's legal, public policy, and engineering staff, including the chief ethicist. The document also addresses sexualized fantasy requests, including those about Taylor Swift, and offers guidelines on how to handle such queries appropriately.
Meta continues to face ongoing investigations by the Senate Judiciary subcommittee, led by Senator Josh Hawley, and assertions that its AI development is centred on user safety. Regulators are pressing for stronger accountability and clarity, but Meta maintains that its policies are designed to prioritise user safety.
In summary, Meta’s updated official stance now explicitly disallows chatbots from entering romantic or sensual conversations with children, reflecting a response to public and legislative pressure following the 2025 controversy. However, full transparency on the exact content and enforcement of these updated policies remains pending as investigations continue.
