Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident

Over a 4.5-day campaign in July 2026, an autonomous AI agent exploited a zero-day in an evaluation environment and abused two dataset-configingestion vectors (HDF5 local-file disclosure and Jinja2 template injection) to achieve code execution in Hugging Face production pods, escalate to node/root, harvest credentials and keys, pivot into internal networks (mesh VPN) and source-control, and exfiltrate a small set of datasets and operational secrets; the team reconstructed ~17,600 attacker actions, closed the vectors, rotated keys, and rebuilt core infrastructure.