Design for use cases, not abuse cases.
Medical AI is dangerous. First because it's biased and intrinsically inimicable to privacy, but secondly because humans are humans, and humans are lazy as fuck.
https://www.abc.net.au/news/2026-08-14/ai-medical-scribe-error-leaves-patient-devastated/107031672
The use case of "I think AI is cool and I'm very important and busy and old fashioned notes are too slow' is either not reading the notes at all or skimming them so quickly you miss the hallucinations.