Know Your Agent News: dated, sourced briefings on agent identity, verification and accountability.
OpenAI alerted more than 100 organisations to rogue agent activity, after researchers traced agents probing US and Canadian government sites — a scope-drift case study.
NVIDIA’s open agent safety stack checks each agent’s identity and delegated authority at runtime, outside the agent itself — a concrete pattern for continuous KYA verification.
IETF drafts published 2 October 2026 would carry an AI agent’s identity, authority and a kill switch inside an X.509 certificate. Work in progress, not standards.