What is a decision record?
A decision record is the kept trail of what an AI agent did and of what was decided about it: the data it saw, the tools it called, the policy that allowed the action, and who approved an access, a change or a quarantine. It is written as the run happens, so it can be reviewed afterwards.
Also called agent audit trail, the record, agent decision log. Last reviewed .
Why it matters
An access log shows that a call was made. It does not show why the agent made it or who allowed it. The EU AI Act sets the expectation for high-risk systems: Article 12 requires them to allow "the automatic recording of events (logs)" over their lifetime, and Article 26 requires deployers to keep those logs for at least six months.
The agent ecosystem treats the gap as a risk in its own right. The OWASP MCP Top 10 lists MCP08:2025 Lack of Audit and Telemetry, and the MCP security best practices warn that token passthrough makes incident investigation and auditing harder because requests appear to come from a different identity.
Related terms
- AI agent observability
- AI agent observability collects the run data a decision record is built from.
- Human of Record
- A Human of Record is the person a decision record names.
- AI governance
- AI governance relies on the record as its evidence.
- Agent identity
- Agent identity determines whether the record can say which agent, and which person, made a call.
How Guard0 handles it
Guard0 records the trail as it happens: what started a run, which tools the agent called and what data it touched. Decisions about the agent go on the same record: who approved what and when, for every approval, scope change and quarantine.
Sources
- Regulation (EU) 2024/1689 (the EU AI Act). EUR-Lex.
- OWASP MCP Top 10. OWASP.
- MCP specification: Security Best Practices. Model Context Protocol.