Psychedelic therapy trials
Psilocybin and MDMA, given with therapy, show promising effects in trials but have not yet been approved by the FDA.
Open in the interactive tree →Substances such as psilocybin and MDMA were banned in the 1970s but returned to clinical research in the 2000s. In one or a few sessions with guided psychotherapy they appear to reduce treatment-resistant depression and PTSD, possibly through a lasting increase in brain plasticity. Open questions include placebo effects (patients know they took the drug), safety and the cost of therapist time.
As of October 2026
The FDA declined to approve MDMA-assisted therapy for PTSD in August 2024 after an advisory panel voted 9 to 2 that the data did not show effectiveness. Compass Pathways reports positive results from two phase 3 trials of its psilocybin product COMP360 in treatment-resistant depression (COMP005 and COMP006, over 1,000 participants); it holds FDA breakthrough-therapy status and a national priority voucher, the FDA agreed to a rolling submission in April 2026, and the company expects to finish the application in the fourth quarter of 2026. There was no approval as of the company's August 2026 report, and it targets a launch in the first half of 2027. Australia has allowed authorized psychiatrists to prescribe psilocybin and MDMA since July 2023.
Open steps
- Trials that stay blinded Low AI leverageMore than 90% of participants in high-dose psilocybin trials guess their treatment, so effects may be inflated; better placebos and masked raters are needed.
- Plasticity drugs without the trip High AI leverageNon-hallucinogenic analogs acting on the same receptor could skip long supervised sessions, but they must still show effects in humans.
- Predicting who responds High AI leverageSome patients respond strongly and others not at all; brain imaging, EEG and session recordings might predict outcomes and guide dosing.
- Delivering therapy at scale Low AI leverageTwo therapists per patient for hours makes treatment costly; shorter sessions, group formats or supported preparation must keep safety.
Where AI could help
Low AI leverage. Approval hinges on trial design, blinding and regulators; AI can predict responders but cannot settle efficacy.
- Predicting who responds from session recordings and baseline data
- Designing analogs without hallucinogenic effects
- Analyzing session audio and imaging to standardize therapy
- Monitoring patients between sessions
Shown so far
- In August 2023 Psychopharmacology published a model that predicted COMP360 psilocybin outcomes from recorded integration sessions of 101 trial participants with 85-88% cross-validated accuracy (according to the company, single trial). source