How AI Companies are Handling Suicide and Self-Harm Today

Key Findings

• — Identifies six common intervention types companies deploy in response to suicide/self-harm disclosure, ranging from crisis-hotline referral to therapeutic techniques such as grounding
• — Organizes implementation challenges into seven categories: cross-cultural content detection, risk assessment for vulnerable populations, balancing validation against reinforcing harmful thinking, designing effective human-care handoffs, applying clinical guidance built for humans to AI systems, managing conflicting cross-jurisdictional regulatory requirements, and measuring meaningful outcomes
• — Synthesizes workshop input from frontier AI company representatives alongside PAI's independent analysis of current practices

Methodology Notes

Based on a March 2026 PAI-hosted multistakeholder workshop (frontier AI companies, mental-health clinicians, researchers, lived-experience contributors) combined with PAI's independent review of company practices; not a controlled study. A downloadable full-length version was referenced on the page but not resolved during verification.

Sources