The OpenID Foundation’s AI Identity Management Community Group (AIIM CG) has published a paper, “A Protocol-Agnostic Threat Classification for Cross-Boundary AI Agent Authorization” (announced October 12, 2026). The paper catalogs threats that arise when AI agents cross security boundaries between platforms: verifying which agent is actually acting, how domains establish trust in each other’s keys and assertions, agents acting on the wrong resource after crossing a boundary, permissions being broadened when translated between domains, access outliving the task it was granted for, and credential capture and reuse. It argues that treating an agent as if it were a human user is the common root of these threats, and that agents run familiar federation threats at machine speed. The document is positioned as a threat model and shared baseline, not a solution, with follow-up work planned on authentication, authorization, and mitigation guidance. Recommended actions include mapping every trust boundary agents cross, investing in authorization-time controls rather than after-the-fact detection, and treating interoperability as a security control. https://openid.net/oidf-publishes-threat-model-for-cross-boundary-ai-agents/