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

Back to mitigations
Proposed

Grounded, dated medical sources with uncertainty on the surface

No naked imperative. Show the evidence grade and the date.

Who should own it
Labs and medical publishers
Frontier labs
How quickly it can land
Weeks
A dedicated squad can land it inside two months.
Expedited implementation
4 weeks
30 calendar days with a crash team
Normal implementation
4 months
120 calendar days as a planned program

Risk this mitigates

20
De facto clinical AI without clinical controls

Given that a general chatbot is already a weekly health advisor for hundreds of millions of people, there is a possibility of medical decisions being made on model output that is not a licensed device, not a record, and not under a clinician’s duty resulting in population-scale diagnostic and dosing errors, and delayed care for people who trusted the chat more than a clinic.

Residual composite 20 · still above the threshold

Effect if implemented

Applied to every failure scenario on that risk, then re-ranked. Axes are clamped at 1.

Likelihood
0
Consequence
1
Urgency
0