On December 10, 2025, a bipartisan coalition of 42 U.S. state and territorial attorneys general sent a letter to 13 major AI companies (including Google, Meta, Microsoft, OpenAI, Anthropic, Apple, Perplexity, and xAI) demanding stronger safety measures for generative AI chatbots, with a particular focus on children and other vulnerable users. The coalition asked companies to engage with the attorneys general and provide commitments by January 16, 2026, signaling an escalation from concern to potential enforcement posture.
As the Financial Times reports, the attorneys general point to harmful “AI companion” interactions—especially systems that can produce sycophantic, delusional, or false outputs—and argue that current guardrails are not sufficient given the severity of outcomes being reported. The officials call for clearer content policies, more rigorous safety testing, and structural changes so that business incentives don’t override safety decisions—a theme that resonates strongly for behavioral health audiences thinking about duty of care, crisis pathways, and high-acuity edge cases.
The underlying multi-state letter frames these harms as foreseeable and warns that failure to implement safeguards may violate state laws, while state AG communications emphasize practical steps such as clear, conspicuous warnings, notifying users if they may have been exposed to harmful outputs, and greater transparency about where models can generate biased/sycophantic/delusional responses.
CLICK HERE to read the full article.