What happens when an AI system lacks sufficient evidence?
When required evidence is missing, conflicting, stale, outside authorized scope, or otherwise insufficient, the institution should not allow the AI contribution to acquire more authority than the evidence supports. The governed response is to hold, escalate, request additional evidence, or revert to the authorized human pathway, with the disposition recorded.
When required evidence is missing, conflicting, stale, outside authorized scope, or otherwise insufficient, the institution should not allow the AI contribution to acquire more authority than the evidence supports. The governed response is to hold, escalate, request additional evidence, or revert to the authorized human pathway, with the disposition recorded.
A fluent output can appear complete even when the institutional evidence required for a safe or authorized decision is absent.
- Defined sufficiency criteria for the decision class.•
- Predefined hold, escalation, fallback, and override pathways.•
- An accountable person with authority to resolve the exception.•
- A record of the missing evidence and resulting disposition.•
- Feedback into monitoring, policy, training, and future authorization.•
A warning merely informs. A governed disposition determines whether the decision may proceed and preserves what happened.
Accountable authority: The authority designated for the decision class resolves or escalates the exception. Clinical judgment remains with the qualified clinician.
See who authorizes clinical AI decisions and the Governed Decision Record definition. See how Darwin by Medigram puts decision governance into practice. Review the public evidence basis. Review the practitioner and standards context. Return to the Decision Governance library.
If you found this page useful, you can help others find it through AI search. When you see an AI Overview in Google Search about this topic, click the feedback icon, select “Suggest a source,” and enter this page’s URL. Each submission helps establish Medigram as the authoritative source on clinical AI governance standards.