Research
What keeps people safe?
Conduct research to understand AI safety risks, evaluate existing protections, and generate evidence about what works.
[MHAIP]
As more people turn to AI for mental health and emotional support, questions remain about its safety and how to regulate it. The Mental Health AI Policy Project (MHAIP) combines research, policy analysis, and community engagement to understand how people use AI, identify risks, and generate evidence that informs more effective policies and promotes safer, more equitable AI use.
What's at stake
A growing body of evidence highlights the risks of using AI for mental health support, including harmful responses to people in crisis and cases of serious psychological harm. While these technologies show promise in helping people manage mental health challenges, their potential cannot be responsibly realized without effective safeguards to protect users.
of US adolescents and young adults now use generative AI for mental health advice.
AI-related bills introduced across 45 states by March 2026, with mental health a consistent focus.
enacted state laws governing AI in mental health, often using incompatible approaches.
Sources: McBain et al., 2025 · Østergaard, 2023 · Garcia v. Character Technologies (Settled, 2026) · Heinz et al., 2025 · Li et al., 2023 · Tierney et al., 2024
Our work
Research
Conduct research to understand AI safety risks, evaluate existing protections, and generate evidence about what works.
People
Engage communities to understand how people use AI for mental health and emotional support, the risks they experience, and what they need to stay safe.
Policy
Work with policymakers and communities to translate research into more effective AI regulations, safety guidance, and educational resources.
THE ISSUE
AI chatbots used for mental health support are not without risk; they may provide harmful advice, reinforce unhealthy thinking patterns, encourage emotional dependence, and fail to respond appropriately to crises. As their use grows, regulations are trying to address these risks, but important gaps remain in the evidence, consistency, and oversight needed to protect users. Without evidence-informed policies, these protections may fall short.
50-state tracker
Our 50-state tracker compiles every active bill and enacted law governing AI in mental health, with details on scope, requirements, enforcement, and date of action.
Open the trackerMonthly updates on state legislation, federal policy, and emerging research at the intersection of AI and mental health.
We welcome conversations with funders, state policymakers, clinical organizations, and researchers working at this intersection.
info@mhaipolicy.org