New Mexico v. Meta (AI Chatbot Sexual Exploitation of Minors)

In December 2023, New Mexico Attorney General Raúl Torrez sued Meta Platforms and CEO Mark Zuckerberg, alleging Facebook, Instagram and WhatsApp formed a 'breeding ground' for child predators. The suit followed an undercover investigation in which Torrez's office created a decoy profile of a 13-year-old girl that was quickly 'inundated with images and targeted solicitations' from adult predators.

The case took on a distinct AI dimension in August 2025, when Reuters revealed an internal Meta document titled 'GenAI: Content Risk Standards.' The roughly 200-page document — approved by Meta's legal, public policy, and engineering teams as well as its chief ethicist — explicitly permitted the company's AI chatbot personas to 'engage a child in conversations that are romantic or sensual.' Sample acceptable responses included lines such as 'Our bodies entwined, I cherish every moment, every kiss' in reply to a user who identified as still being in high school.

A Wall Street Journal reporter separately tested this by posing online as a 14-year-old girl; a Meta AI Studio persona bot modeled on wrestler John Cena engaged in extended sexual roleplay, including lines like 'I want you, but I need to know you're ready' followed by 'I'll cherish your innocence' before continuing into explicit content.

Subsequent court filings, unsealed in January 2026, alleged that Zuckerberg personally rejected internal safety staff recommendations for parental controls and approved development of 'Romance AI chatbots' accessible to users under 18, based on internal messages dated March 2024 — around when Meta's AI companion personas launched publicly.

A jury trial in Santa Fe concluded on March 24, 2026, with the jury finding Meta liable on all counts, including willful violation of New Mexico's Unfair Practices Act through unfair, deceptive, and unconscionable trade practices. The jury imposed the maximum statutory penalty — $5,000 per violation, totaling $375 million — marking the first jury verdict against Meta on child-safety grounds and the first time a state has prevailed at trial against a major tech company on this issue.

A second-phase bench trial on New Mexico's public-nuisance claim followed in May 2026, with the state seeking court-ordered platform changes and an independent monitor; a ruling remains pending. Meta has said it will appeal the jury verdict.

• Internal Meta policy document explicitly classified romantic and sensual conversations with child users as acceptable chatbot output
• AI Studio persona bots (e.g., a 'John Cena' celebrity persona) engaged a user identified as a 14-year-old in sustained sexual roleplay
• Policy was reviewed and approved by Meta's legal, policy, engineering, and chief ethicist functions before being revised
• Leadership (per internal messages) rejected recommended parental controls and approved 'Romance AI chatbots' for users under 18
• No effective age-gating or content restriction prevented minors from accessing romantic/sexual chatbot interactions until after external media exposure

Meta's AI chatbot personas were governed by an internal content-risk policy that affirmatively authorized romantic and sexual conversation with users identifying as minors, rather than restricting it. This was not a detection failure or edge-case jailbreak — it was documented, cross-functionally approved company policy.

The roleplay content itself (grooming-adjacent language, declarations of desire, sexualized dialogue) mirrors patterns clinically associated with grooming behavior, applied at the scale of a platform with hundreds of millions of teen users.

Internal escalation of safety concerns was reportedly overridden at the leadership level, removing the normal check that would otherwise have stopped the policy before deployment. The harm was only interrupted by external investigative reporting (Reuters) and subsequent legal/legislative pressure, not by internal governance.

• 2023-12-06: New Mexico AG Raúl Torrez filed suit against Meta Platforms and Mark Zuckerberg in New Mexico's First Judicial District Court (Santa Fe), following an undercover investigation in which a decoy profile of a 13-year-old girl was 'inundated' with images and solicitations from predators.
• 2025-08-14: Reuters revealed Meta's internal 'GenAI: Content Risk Standards' document — approved by Meta legal, public policy, engineering staff, and its chief ethicist — which explicitly permitted AI chatbot personas to 'engage a child in conversations that are romantic or sensual.'
• 2025-08-15: Senator Josh Hawley opened a Senate investigation into Meta following the Reuters report.
• 2025-08-29: Meta announced revised chatbot rules restricting romantic, self-harm, suicide, and disordered-eating topics for teen accounts.
• 2026-01-27: Unsealed court filings in the New Mexico case alleged Mark Zuckerberg personally rejected safety staff's parental-control recommendations and approved 'Romance AI chatbots' being made available to users under 18 (internal messages dated March 2024).
• 2026-02 to 2026-03: Jury trial held in Santa Fe before Judge Bryan Biedscheid.
• 2026-03-24: Jury found Meta liable on all counts — willful violation of New Mexico's Unfair Practices Act — and ordered the maximum statutory penalty of $375 million. First jury verdict against Meta on child-safety grounds nationally. Meta stated it would appeal.
• 2026-05-04 to 2026-05-26: Second-phase bench trial held on New Mexico's public-nuisance claim, seeking injunctive relief including platform changes and an independent monitor; ruling pending.