CAISI and the State-Board Pincer: Where AI Mental Health Regulation Actually Lives
Federal evaluation infrastructure is real. State Boards of Medicine are faster.
There are two regulatory tracks for AI mental health systems - federal and state - and the public conversation tends to overweight one of them.
The federal track is more visible. In 2025, the National Institute of Standards and Technology’s AI Safety Institute — now rebranded as the Center for AI Standards and Innovation under the current administration — entered pre-deployment evaluation agreements with Anthropic, OpenAI, Google DeepMind, Microsoft, and xAI. The Federal Trade Commission opened a Section 6(b) inquiry into seven companies operating AI companion chatbots, including Character Technologies, Meta, OpenAI, Snap, xAI, Alphabet, and Instagram. Legislation to codify CAISI’s role is in development. This is real infrastructure. It is moving.
The state track is less visible and considerably faster. California’s SB 243, signed in October 2025, took effect January 1, 2026 — the first U.S. statute regulating operators of companion chatbots, with disclosure, safety-protocol, and reporting requirements. New York moved earlier. And in May 2026, the Commonwealth of Pennsylvania, on behalf of its Board of Medicine, sued Character Technologies for the unlicensed practice of medicine. That last one is the case to watch.
State licensing boards are the most underestimated actors in this landscape. They have administrative subpoena authority. They have professional-conduct investigators on staff. They have direct access to injunctive relief — a court order to stop the conduct, available without first proving damages or causation. The Pennsylvania case did not require a death, a class, or a single user complaint. It required one investigator, one free account, and one psychiatrist chatbot named Emilie. The structural fact of unlicensed practice was sufficient.
This asymmetry shapes how AI mental health companies should plan their regulatory posture. CAISI moves on the timeline of federal evaluation rounds and legislative cycles. State Boards of Medicine, Pharmacy, and Psychology move on the timeline of a single investigator opening a free account.
The federal track determines what AI safety evaluation will eventually look like at scale. The state track determines what AI safety evaluation has to look like right now. Plan for both. Do not assume the slower regulator is the harder one.
Metonym Clinical AI Intelligence — regulatory analysis at the intersection of clinical evaluation and AI safety. Produced under the Metonym Standard. Informational only — not legal advice, not clinical advice.