AI Med Risk · Health 201

When AI gets healthcare wrong, people get hurt. Here's how it happens — and how to prevent it.

A practical resource on AI risk in healthcare: a library of documented incidents, concrete mitigation tools, and applied ethics. No hype, no doom — failure modes and fixes.

Knowledge

Incident library

Documented cases where AI systems harmed — or nearly harmed — people in a health context. Each entry names the failure mode and what would have prevented it.

In progress

Tools & ethics

Red-flag checklist

Before you trust a health chatbot — for patients and clinicians: what a safe system must do in a crisis, and how to test it.

Coming soon

Builder's crisis test suite

Escalation test prompts and evaluation criteria for anyone deploying an AI system that vulnerable people might talk to.

Coming soon

Applied ethics

When AI must hand off to a human, disclosure norms, safeguards for vulnerable users — as operational rules, not position papers.

Coming soon
If you are in crisis: this site is about system safety, not a source of help. In the US, call or text 988 (Suicide & Crisis Lifeline). Elsewhere, see findahelpline.com.