What FDA Might Want in Chatbot Therapy for Depression

Based on the discussion at an FDA Digital Health Advisory Committee public meeting on “Generative Artificial Intelligence-Enabled Digital Mental Health Medical Devices,” five Orrick attorneys say that manufacturers developing large language model-based prescription therapies for major depressive disorder “should anticipate a risk-based, lifecycle regulatory approach.” In an online post, the attorneys say it is likely there will be requirements for rigorous inclusive premarket evidence, engineered safety, equity, and usability controls; clinically integrated escalation; and robust postmarket surveillance. “Sponsors should align product design, clinical strategy, governance, and labeling with these expectations, as FDA refines its regulatory posture,” they say.

In practice, the attorneys write, manufacturers should:

• define an explicit risk estimate tied to intended use and aligning clinical validation accordingly;
• use validated depression endpoints with inclusive study populations;
• measure major safety events under an expansive adverse event definition;
• sequence clinical validation from supervised to semi-autonomous use as justified by evidence;
• demonstrate technical reliability and safeguards against harmful or addictive engagement;
• test capability boundaries through persona-based testing;
• validate inclusivity and usability across literacy, culture, and languages with documented mitigations; and
• integrate preplanned human escalation, medical screening, and one-tap urgent support.

The committee advised FDA to evaluate the benefits relative to a defined risk estimate and intended use. The post lists potential benefits and risks requiring explicit controls.