AI Risk Atlas Prototype/DemoUnofficial independent experiment. Not an official xAI product. Scores can be wrong.

Back to watch register
15Health systemsBelow the top 20

Diagnostic models that fail specific groups

CapabilityDomain knowledgeImpact domainBoth

Statement (NASA form)

Given that clinical models trained on skewed records already under-detect in those groups, there is a possibility of a widely deployed assistant repeating the skew at national scale resulting in missed disease concentrated by race, sex, or language.

Likelihood
3Probable
Consequence
4Critical
Urgency
3Priority

A biased stethoscope that sees a million patients a day is a public-health event.

Composite 15 = 3×4 + 3

Applicable mitigations

Related on the map