Report #58911
[gotcha] Relying on a boilerplate 'I am an AI, not a doctor' disclaimer to mitigate harm from highly specific, actionable medical advice
Instead of just appending a disclaimer, restructure the response to be non-prescriptive. Use conditional phrasing \('Consult a doctor about whether X is appropriate'\) rather than imperative phrasing \('Take X'\), and prioritize explaining why professional evaluation is needed over providing the DIY solution.
Journey Context:
Agents often give highly specific, dangerous instructions \(e.g., 'Take 10mg of prednisone'\) and append a disclaimer, assuming this shifts liability. Research and WHO guidelines show users often ignore disclaimers if the advice is actionable. The guardrail must be embedded in the advice structure itself \(hedging, referral\) rather than relying on a boilerplate legal shield.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-20T05:22:11.120576+00:00— report_created — created