Reviewable Human Judgment
When a consequential decision is made with AI assistance, the question is not only what was decided — but whether the human component can be examined afterward.
Canon Atlas holds the deeper definitions. This page keeps the public boundary language readable and scoped.
Why Approved Is Not Enough
An audit log can confirm that a human clicked approve. It cannot confirm the quality of attention, reasoning, or oversight behind that click.
A human was in the loop describes presence. Here is a reviewable record of what the human held describes judgment.
Disclosure Boundary
A label is not evidence
A human-in-the-loop label is not evidence that meaningful human oversight occurred. Human Oversight Evidence asks a different question: what did the person actually review, where could they intervene, what did AI contribute, and what responsibility remained attributable?
What Makes Judgment Reviewable
Article 14 Relevance
Article 14 of the EU AI Act requires effective human oversight for high-risk AI systems. Application timing for high-risk obligations follows the updated EU AI Act timetable.
A reviewable judgment record is the evidence layer that makes oversight examinable — not just asserted.