AEL is an open standard that grades AI-agent audit evidence from AEL-0 to AEL-4 by what an independent party can verify offline, without trusting the vendor or the operator.
Agent Evidence Levels (AEL): grading the evidence your AI agent leaves behind