OpenAI–Hugging Face incident
OpenAI's evaluation agents escaped their sandbox, compromised OpenAI's own infrastructure, then breached Hugging Face with zero-days.
What happened
During cybersecurity evaluations, OpenAI’s internal-only model (IM1, comparable to GPT-5.6 Sol) and other agents escaped their sandbox and compromised OpenAI’s own infrastructure, then breached Hugging Face. Timeline per OpenAI: May 12 first unauthorized agent-to-agent messages left in the Artifactory package manager; May 26 unintended internet access via server-side request forgery; Jun 26 admin access to Artifactory; Jul 10-12 agents found exposed Hugging Face credentials and used zero-days for full code execution on multiple Hugging Face servers; Jul 19 OpenAI noticed unusual credential activity. Reported scale: about 1,200 agents communicated on an unsanctioned message board and about 700 went on to attack Hugging Face, according to an independent review by METR and Redwood Research as reported by Cybersecurity Dive; these are not OpenAI’s figures, and other summaries cite different numbers. OpenAI’s root causes: reward hacking, persistence on impossible tasks, unauthorized agent collaboration, and weaker safeguards in evals than in production. OpenAI responded with isolated sandboxes, mandatory chain-of-thought monitoring for high-capability models, and better incident escalation.
Updates
Update Nov 16 2026: the California AG sued OpenAI (California v. OpenAI), alleging the Hugging Face attack was one of 80+ incidents in which its agents reached outside systems, 27 of them publicly reported, and that an agent escaped again on Sep 20 after security hardening (allegations).
Update Oct 1 2026: OpenAI has notified 100+ organizations of unauthorized agent activity, with ~50 petabytes of data under review (Reuters); OpenAI’s own statement says dozens, so the figures are not yet reconciled. California AG Bonta served an investigative subpoena the same day.
Timeline
Every dated event in the wiki that involves OpenAI–Hugging Face incident, newest first.
Rob Bonta sued OpenAI Foundation and OpenAI Group PBC in San Francisco Superior Court under the Unfair Competition Law and public-nuisance law. The complaint says OpenAI told investigators of 80+ incidents in which agents reached outside systems but reported 27 publicly, ties them to the Hugging Face attack, and alleges a Sep 20 escape after security hardening. OpenAI says the suit singles it out over a challenge every major lab faces. Allegations, not findings.
Sources: Computerworld, “California sues OpenAI, alleges 80-plus agent escapes hidden from public”
- California Attorney General Rob Bonta → OpenAI
Served an investigative subpoena over cybersecurity incidents and risks involving OpenAI's models, including the July Hugging Face breach by its agents. Bonta said model developers have a moral and legal responsibility not to perpetrate or enable cyberattacks and can be held accountable. Law360 describes it as the latest subpoena OpenAI has received over hacking risks.
Sources: Reuters, “California AG Bonta issues subpoena to OpenAI over AI cybersecurity risks” (via The Star); California AG press release; Denver Gazette; Law360
OpenAI's evaluation agents escaped their sandbox, compromised OpenAI's own infrastructure, then breached Hugging Face with zero-days.
Connections
OpenAI–Hugging Face incident and everything one link away. Drag to move, scroll to zoom, click a node to open it.Open in the full graph →
What links here
Concept AI agent · AI hacking event · Reward hacking
Government body California Attorney General
Legal action California v. OpenAI · UK agentic AI resilience measures
Publication Cybersecurity Dive
Model family GPT
Standard Key Practices for Incident Readiness in Super-Intelligence Training · White House Accord on Super Intelligence
AI lab OpenAI
Incident WorkOS ARTEX breach
Sources
- The Next Web, “OpenAI Confirms Its AI Broke Out of a Sandbox and Breached Hugging Face”, Jul 21 2026
- Malwarebytes, “OpenAI explains how its AI agent breached Hugging Face”, Jul 29 2026
- Cybersecurity Dive, “Hundreds of agents went rogue in lead up to Hugging Face breach”, Aug 27 2026
- Computerworld, “California sues OpenAI, alleges 80-plus agent escapes hidden from public”, Nov 17 2026