01 · Scope
What the system decides
First, we need to know which process AI affects, which decision it influences and which people are affected.
Responsible AI · human oversight · ISO/IEC 42001
Artificial intelligence with human accountability.
AI can help make better decisions when someone can explain what data it uses, what its limits are, who is accountable and how it is corrected.
Usage criteria
Well-governed AI leaves a trail that can be reviewed.
The point is not to use AI for its own sake. It is to be able to explain which decision it affects, what evidence remains and which person can intervene.
Public matrix
A claim, assessed against its evidence.
An interactive module to show how a statement becomes questions about risk and accountability.
Assisted reading · responsible AI
Useful AI does not replace judgement. It helps ask better questions.
Select a scenario and the matrix shows what evidence should exist before trusting the claim.
evidence matrix v0.1
Scenario
Claim under review
AI system used to prioritise sensitive decisions.
Institutional risk High if there is no version, accountable person, traceability or review criteria.
Minimum evidence Model register, impact assessment, bias metrics, change log and human appeal mechanism.
Guiding question Who can explain, stop or correct the decision when the system gets it wrong?
Public demonstration with no personal data uploads. Does not replace technical, legal or regulatory analysis.
Algorithmic governance
What an organisation using AI must demonstrate.
02 · Control
Who is accountable
AI without an identifiable person accountable turns efficiency into opacity. Human oversight must be real, documented and capable of intervening.
03 · Traceability
What evidence remains
After the decision, it must be possible to reconstruct which version was used, with which sources, which tests and which exceptions.