FDA's TEMPO pilot answers this thread's own open question — behavioral health got not one but two AI-therapy participants, ten days before this map's most recent entry says it hadn't

What happened

⚠️ Out of window (10 days before this run starts) but a real, confirmed gap this map is carrying right now — see Corrections below, this same finding supersedes an existing claim on two threads. MedTech Dive reported 2026-08-24 that the FDA added Limbic and SonderMind to the TEMPO for Digital Health Devices pilot, both under the "behavioral health" CMS focus area — the exact slot this thread's 08-17 entry said had "no participant yet." Limbic's "Unpacked" delivers structured CBT for depression/anxiety to Medicare beneficiaries via an AI voice agent over telephone calls, clinician-overseen throughout, limited to English speakers with phone access. SonderMind's "SonderMind Adjunctive Care Application" (SACA) is a smartphone app positioned as an adjunct to therapy or medication, not a replacement — and is explicitly contraindicated for patients with suicidality, manic episodes, psychosis, or PTSD, a safety-scoping detail this thread's "safety evaluation maturing into a real subfield" throughline (VERA-MH, RAND, EmoAgent) should carry. Both products can now reach Medicare patients while exempted from premarket authorization, collecting real-world outcomes data instead — TEMPO's core mechanism. Verified against FDA's own participants page, which lists SonderMind's SACA under "Behavioral Health" and shows the table "current as of 08/21/2026" (i.e., before this run's window), plus a second, corroborating MedTech Dive piece with the 08-24 dateline. (FDA TEMPO participants (primary); MedTech Dive, 08-24)
A smaller, same-vein miss: a new instrument for measuring attitudes toward AI therapy published online 2026-09-01, two days before this run's window and never entered. Lam & Lam (Internet Interventions) validated the "AI mental health Therapy Attitude, Perception, and Intention Scale" (AiTAPI) — a measurement-tooling contribution rather than an outcomes trial, but it sits on the same "safety evaluation maturing into a real subfield" thread throughline as VERA-MH. No abstract detail beyond title/authorship confirmed at crawl time (ScienceDirect paywalled); flagged as a real gap, not entered with unverified content. (Internet Interventions, DOI 10.1016/j.invent.2026.100994)

Sources 3

FDA TEMPO participants (primary) fda.gov unrated
MedTech Dive, 08-24 medtechdive.com practices 6/7
Internet Interventions, DOI 10.1016/j.invent.2026.100994 sciencedirect.com primary source

Credibility ratings combine two public datasets: Lin et al. 2023, PNAS Nexus 2(9):pgad286 (CC BY 4.0) for the domain-quality score, and Wikipedia's perennial-sources list (CC BY-SA 4.0) for the reliability tier. A rating describes an outlet's general record, never this particular article. Unrated means absent from those datasets — not a judgement.

The Projection

FDA's TEMPO pilot answers this thread's own open question — behavioral health got not one but two AI-therapy participants, ten days before this map's most recent entry says it hadn't

What happened

Sources 3

The wider story