California spreads responsibility for clinical AI: hospitals and clinics will be held accountable, not just developers

By Law360 · November 4, 2026.

The California Legislature has decided that hospitals and clinics, and not just artificial intelligence developers, will be held accountable for how AI behaves at the patient's bedside. That is how Paul Pitts and James Hennessy sum it up in an analysis piece published in Law360, which describes new state laws setting ground rules for AI in clinical care.

A note on the material: the text we were able to access is restricted by Law360's paywall and shows only the lede. We do not know the provisions, effective dates, specific obligations or penalties. For that reason, we will not detail requirements we cannot verify. The article's tags list the Medical Board of California among the related bodies, but we cannot specify what role it would play.

What is clear is the shift in approach. Until now, much of the debate over liability in healthcare AI revolved around the model's maker. Under this reading of the law, the facility deploying the tool becomes jointly responsible for its use on patients.

**Our take.** We think this is a sensible move. Responsibility should lie where real control lies: the hospital decides which tool it adopts, with what data, under what oversight and in what workflow. A developer cannot monitor every local deployment. If liability rests only with the vendor, the facility has little incentive to audit, train its staff or keep a human in charge.

This approach fits an idea we often advocate: the value and risk of an AI system lie in the deployment layer, that is, in who acts, with what permissions and who can correct it, more than in the model in the abstract.

There are also short-term risks. A heavy compliance burden could hold back small and rural hospitals, which have fewer legal and technical resources. That would widen the gap with large health systems. In addition, excessive legal caution could delay useful diagnostic tools. The more precise the rules, the less temptation there will be to regulate panic instead of actual capability.

In the long run, healthcare is probably the field where AI contributes the most: early diagnosis, drug discovery and, eventually, diseases that cease to be diseases. Getting there requires trust, and trust is built with clear liability rules. In general, California tends to set the regulatory tone that other states and companies end up adopting, so it is worth following how this law takes shape once its full text is published.