Evidence type
Runtime evidence
Runtime evidence is valuable. It can show what an AI agent, model gateway or system action did, what policy was applied, whether an action was allowed or denied, and whether a record has been tampered with.
Useful, but not the whole decision
Runtime records can show what happened in a system. They can preserve action order, policy outcomes, system state and integrity signals that are important to later review.
But runtime evidence is not the same as decision evidence. A runtime record may not identify the source evidence considered, the business authority in force, the governing conditions, the human judgement or the organisational commitment that made the action consequential.
A signed runtime record may become an Evidence Object inside a decision-evidence package. It does not, by itself, establish the complete business decision-state unless it is connected to the relevant evidence, authority, governing conditions, human judgement and organisational commitment.
Questions for reviewers
- What decision did this runtime action contribute to?
- Which conditions and authority applied at the execution boundary?
- Was the action available, relied upon, excluded or superseded evidence?
- What remains unknown beyond the runtime record?