Photo: Google DeepMind / Pexels
The FDA has cleared over 1,500 AI-enabled medical devices, but not a single generative AI product has been authorized for mental health treatment.
The US government is pushing AI into healthcare faster than the guardrails can keep up. Federal agencies are embracing artificial intelligence tools to address a deepening mental health crisis, even as regulators acknowledge they haven’t figured out how to ensure these systems are safe for some of the most vulnerable patients.
The FDA has authorized 1,524 AI-enabled medical devices as of August 2026. Zero of those clearances cover generative AI products designed to manage mental health conditions.
The federal push for AI in mental health #
Dr. Mehmet Oz, serving as CMS Administrator, publicly endorsed the use of AI avatars for mental health intake in rural areas back in February 2026. His argument was straightforward: staffing shortages in rural mental health care have gotten so bad that an AI chatbot might be better than no therapist at all.
The FDA, meanwhile, launched its TEMPO pilot project in late 2025 to evaluate AI tools in real-world clinical settings. The program is specifically designed to collect data on how effective these tools are at treating mental health conditions, with a particular focus on depression. Manufacturers for the pilot are expected to be selected by September 2026.
The risk profile regulators can’t ignore #
An FDA Digital Health Advisory Committee meeting on November 6, 2025, laid out the tension in stark terms. Committee members discussed the potential benefits of generative AI for broadening access to mental health care while simultaneously flagging a list of risks: model drift, embedded biases, privacy vulnerabilities, and the danger that patients might place too much trust in AI systems that lack the clinical judgment of a human therapist.
States aren’t waiting for Washington #
While federal regulators deliberate, several states have decided to write their own rules. Illinois and California have both moved to enact laws regulating AI’s use in therapeutic contexts. These state-level efforts generally focus on requiring disclosures to patients about when they’re interacting with AI rather than a human, and mandating safety protocols for AI-powered therapeutic tools. The fact that no generative AI has been sanctioned for therapeutic use by mid-2026 suggests that federal regulators are aware of the gap between what the technology can do and what they can confidently vouch for. The TEMPO pilot’s results, expected to start materializing after manufacturer selection in September 2026, could prove pivotal.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our