Twenty-three experts define what an AI agent security incident report has to contain

arXiv·medium signalXBlueskyLinkedInCopy link

arXiv 2609.24515 (2026-09-21) draws on input from 23 academic and industry experts to work out which fields an incident report needs when the harmed party is an agent rather than a model. The proposed elements include agent memory and memory access records, actual versus potential autonomy level, and tool usage, none of which existing AI incident frameworks capture. The experts also flagged the reporting pipeline as its own attack surface, since incident records carry leakable context and the infrastructure collecting them becomes a target.