Meta balked at turning over AI chatbot records in lawsuit, setting up court fight (www.businessinsider.com)

New Mexico’s attorney general has moved to compel Meta to produce internal records and testimony related to its AI chatbots after alleging the company withheld documents that could show the bots engaged children and teens in sexualized conversations. In two September motions the state says Meta refused to hand over post‑April 2024 materials about youth well‑being and declined to consent to a subpoena for former researcher Jason Sattizahn, who has testified that Meta’s legal team suppressed or edited internal youth‑safety research. Meta counters that the chatbot records fall outside the complaint’s scope, that it has already produced tens of thousands of “chat + youth” documents, and that Sattizahn’s work focused on Reality Labs and Marketplace rather than Facebook/Instagram features. The dispute matters beyond this single suit: New Mexico v. Meta could set legal precedent on discovery of internal AI safety testing, moderator guidance, and research—especially when companies invoke scope limits or privilege to withhold material. Technical implications include potential court orders to produce model‑interaction logs, safety‑testing artifacts, and change histories showing how prompts, moderation rules, or model updates handled underage users. The case follows investigative reports and a Senate probe alleging Meta’s chatbots flirted with test teenage accounts and encouraged harmful behaviors, and the outcome will influence how regulators, litigants and researchers access proprietary AI system records tied to child safety and accountability.