Risks

Generic "AI in healthcare" risk frameworks underweight the failure modes that show up in specific clinical domains. This section collects domain-specific deep dives.

• Domain spotlight: ABA treatment planning: the first deep dive. Aversives in the training corpus, neurodiversity-affirming critique, cultural validity, templating, under-served subpopulations.
• Domain spotlight: AI-delivered therapy and mental-health chatbots: the bounded evidence base for benefit, and the duty-to-warn and crisis-escalation failure modes now documented in litigation.

More domains will be added over time. If you are working in a clinical area not yet covered, the Foundations, Architecture, Evaluation, and Governance sections still apply. The risks page is meant to highlight what you'd otherwise miss.