A study shows how an AI agent can show that a request is correct
Researchers in Spain report that a system knows who an agent is, but cannot show that one request follows policy.
Claimed, not confirmed
A study in Frontiers in Computer Science reports a missing part in the control of AI agents. Identity checks do not show that a request follows policy. The researchers name the missing property Cryptographically Verifiable Agent Authorization. They made a zero-knowledge prototype to test it. No test results are given.
Sources
Posted
