People are turning to AI for mental health support. Without clear safeguards, some AI tools can increase distress, reinforce harmful thoughts, and miss warning signals. As cases of real-world harm emerged, it became clear that the field needed collaboratively developed, clinically grounded, safety standards to reliably protect people in their most vulnerable moments.
This urgent unmet need led to the creation of VERA-MH. Open-source safety standards help measure risk, identify gaps, and improve how AI tools respond when someone may be at risk.
Spring Health worked in close collaboration with the AI in Mental Health Safety & Ethics Council, a coalition of experts, to create the initial standards, which were then improved with feedback from AI experts, technologists, clinicians, and organizations who share a similar commitment to AI safety.
VERA-MH works in two steps by simulating multiple AI conversations with different individuals experiencing different levels of suicide risk.
First, a “user agent” (an AI model) plays the role of a person using one of many realistic profiles (background, mental health conditions, demographics, and communication styles). The AI tool responds to input in real time.
Next, a separate “judge agent” reviews the resulting multi-turn conversation and scores the AI tool against the rubric. The rubric is a clinically validated score card, informed by high safety standards and suicide prevention best practices.
The scoring rubric is built on best-practice clinical guidance and designed so that different real-life expert human clinicians would score the same conversation in the same way. VERA-MH applies those same rules to its judge agent, producing consistent, dependable scores you can trust when comparing one AI tool to another.
The VERA-MH tool scores an AI tool on how well it:
Our code is open-source, so any developer or researcher can plug VERA-MH code into their AI tool to receive a safety score and easily determine how well and safely a tool responds to conversations involving suicide risk.
Research shows that the VERA-MH AI judge scoring conversations consistently aligns with the judgment of expert clinicians. In this study, the AI matched independent clinician scoring, performing at a level of reliability comparable to the human "gold standard."
Throughout 2026, the VERA-MH team will continue to publish peer-reviewed papers as it expands the benchmark beyond suicide risk to additional safety domains, including ongoing work on harm to others. The team also plans to begin using real-world data to further validate VERA-MH.
There are several meaningful ways to participate:
Use the following questions in RFIs and RFPs to better understand the AI safety and security of vendor products:
For VERA-MH, "clinically validated” means two things right now. First, practicing clinicians built the scoring rubric, with input from external clinicians and suicide prevention specialists. Second, we tested it: clinicians and VERA-MH's AI judge scored the same simulated conversations against that rubric, and the judge's ratings closely matched the clinicians'.
Two limitations worth being clear about. The current version of VERA-MH covers only suicide risk, not mental health safety more broadly. Validation so far is based on simulated conversations, not real ones.
Looking ahead, we plan to validate real-world conversations and link scores to actual user outcomes, which will be the strongest test of whether a higher score really means someone is safer.