AI Agent Incident Register — legal-analysis corpus of AI agent incidents (snapshot 2026-07-22)

A numbered public corpus of AI agent incidents, each analysed legally: the legal duty engaged, who bears liability across the chain, the governance that would have prevented it, and the mapped controls. Authored by Michael K. Onyekwere (CIPP/E, common law qualified lawyer practising as a Data Protection Officer). The distinctive contribution over security-side incident trackers (the AI Incident Database, METR, Oso, MITRE ATLAS) is per-incident legal-liability allocation: every entry carries a liability locus — deployer, shared, or vendor — the modal allocation of who most likely carries the legal exposure. Each entry maps to the OWASP Top 10 for Agentic Applications 2026, the NIST AI Risk Management Framework, Singapore's IMDA Model AI Governance Framework for Agentic AI, and the EU AI Act. Live corpus: https://companyscope.io/register . Machine-readable feed (authoritative): https://companyscope.io/api/register . This file is a point-in-time snapshot (7 entries, AIR-2026-001 through AIR-2026-007) for citation and archival. Legal analysis of public facts, not legal advice.