What happened
On 6 November 2026 (in-window), ISO/IEC JTC 1/SC 42 progressed a new approved work item, ISO/IEC AWI TR 27258 'Artificial intelligence — Interactions in multi-agent AI systems', through the proposal stage (10.99 'New project approved') into the preparatory stage (20.00), both dated 2026-11-06 per the ISO catalogue. The document scope covers interactions in multi-agent AI systems operating in digital, physical and hybrid environments, within or across organizational boundaries.
Why it matters
This is a direct standards-body response to agentic AI: it signals that ISO/IEC is now formalising guidance on how multiple AI agents interact, which is the layer where agent identity, delegation, tool-call and authorization boundaries get defined. It maps to the same agentic-AI requirements (identity, least privilege, auditability) that NIST, OWASP, the IETF and regulators have been racing to cover. Trackers should note the direction and prepare to comment once the draft circulates for ballot.
Action needed
Monitor the ISO/IEC JTC 1/SC 42 work item; register for notification so you can comment on the future CD/DIS draft. Map existing agent-security controls (agent identity, delegation, tool authorization) against the interactions it defines.