On November 6, the FDA’s Digital Health Advisory Committee will meet to talk about generative AI-powered mental health tools. The committee’s job is to enable innovation in this needed space while reducing risks to users. You can read their official statement in the Federal Register.  They’ll also talk about what kind of evidence they will expect to see before approving these tools and on an ongoing basis once they’re on the market. 

The meeting will be held virtually. The FDA says it will post the webcast link and background materials at least two business days before the meeting. Anyone can watch, and anyone can also file a public comment until December 8 through regulations.gov (docket number FDA-2025-N-2338).

Why this meeting matters

There’s been an explosion of AI tools promising mental health support — from chatbots to digital therapy assistants. Many aren’t FDA-regulated, and some are not backed by strong clinical evidence. Also raising the ante on these tools is their focus on direct-to-patient interactions, where no licensed provider oversees the advice and coaching generated by these products. The use of generative AI in these devices will be a major focus given their potential for unpredictable responses.

The FDA will be grappling with big questions like:

• How do we know these tools are safe and effective?
• What counts as enough proof before launch?
• How should companies keep track of risks once people start using them?

And all of this is happening against the backdrop of a real mental health crisis in the US: more people need help than providers can supply. These tools have been celebrated as a potential solution, assuming they’re safe and trustworthy.

Recent flashpoints

The FDA’s decision for the committee meeting was most likely prompted by a few recent events, including the OpenAI lawsuit following the tragic death of 16-year-old Adam Raine, where ChatGPT allegedly gave gave harmful and even encouraging responses around suicide. This case flags the risk of generative AI when used in direct-to-consumer applications, no doubt. 

Beyond this though, the FDA needs to re-up its guidance on SaMD and digital health products. It’s earlier attempts — like the Pre-Cert program — didn’t keep up with how quickly AI has changed. Generative AI wasn’t even on the horizon when those frameworks were designed.

Finally, the lack of clear guidance on what the FDA expects in terms of evidence and post-market monitoring has causing even major players to pause or seek alternatives to product approval. Most notably, Woebot, once given Breakthrough Device designation by the FDA, shut down its chatbot after prolonged back-and-forth with FDA regulators, which was a bad omen for others wanting the pursue FDA approval.  

What to Expect

At this meeting, the FDA will start sketching out how it might regulate AI-powered mental health tools going forward. Expect the committee to weigh in on:

• Premarket evidence requirements, specifically what level and type of clinical data FDA is expecting. 
• Postmarket monitoring guidance regarding how should companies track safety and effectiveness once these tools are in use, e.g., real-world performance monitoring, patient registries, etc.
• What risk mitigation strategies should companies have in place, especially with regard to generative AI, biased algorithms, safeguarding against harmful outputs, etc. 
• Keeping direct-to-patient use safe by potentially flagging high-risk scenarios.

I will definitely follow up with a report out of the meeting.  It should be a juicy one!