{
 "license": "CC-BY-4.0",
 "attribution": "Fide AI, Agentic Cyber Explorer",
 "url": "https://agentic-cyber-explorer.pages.dev/events/openai-control-room-inference-shutdown-2026/",
 "asOf": "2026-11-17",
 "id": "openai-control-room-inference-shutdown-2026",
 "date": "2026-11-09",
 "datePrecision": "day",
 "title": "OpenAI Control Room revokes further agent inference without terminating workload code",
 "lane": "defense",
 "kind": "framework",
 "summary": "OpenAI describes Control Room, infrastructure that registers research jobs against approved safety cases and can revoke their access to further model output. OpenAI reports a maximum 35-second delay from receiving a stop instruction to halting output. Workload code can remain running; the mechanism does not stop every action an agent has launched.",
 "whyItMatters": "It makes a concrete shutdown boundary inspectable and distinguishes revoking inference from terminating processes or detecting misconduct.",
 "actors": [
  "openai"
 ],
 "topics": [
  "monitoring-and-control",
  "sandbox-containment",
  "multi-agent-security"
 ],
 "atlas": [
  "monitor",
  "model",
  "human-approver"
 ],
 "artifacts": [
  "control-room"
 ],
 "sources": [
  {
   "url": "https://openai.com/index/control-room/",
   "publisher": "OpenAI",
   "title": "Control Room",
   "date": "2026-11-09",
   "type": "primary",
   "accessed": "2026-11-15"
  }
 ],
 "keyFacts": [
  {
   "fact": "Research jobs register with Control Room against an approved safety case. Inference engines check authorization when fulfilling requests; the mechanism is enforced in the inference layer rather than relying on workload code.",
   "locator": "How it works"
  },
  {
   "fact": "OpenAI reports a maximum 35 seconds between Control Room receiving a stop instruction and further model output stopping, due to cached authorization. This excludes detection and the decision to intervene.",
   "locator": "Speed and reliability"
  },
  {
   "fact": "Workload code continues running after inference is revoked. Code or external actions already launched are outside the stated inference-stop guarantee.",
   "locator": "How Control Room works: Model output stops—the job stays"
  },
  {
   "fact": "At publication, fail-closed authorization was deployed for all frontier-model research workloads, with rollout to other research workloads planned. Monitoring and the decision to stop a job remain separate from inference enforcement.",
   "locator": "Speed and reliability"
  }
 ],
 "significance": 4,
 "fideQuestions": [
  "FID-074",
  "FID-087"
 ],
 "methods": [
  "ai-monitoring",
  "capability-restriction"
 ],
 "review": "assistant-drafted",
 "addedOn": "2026-11-15",
 "revisions": [
  {
   "id": "scope-authorization-method",
   "on": "2026-11-17",
   "kind": "correction",
   "why": "Correction: tagged Control Room for inference authorization and monitoring. The approved safety case applies to the research workload, while inference requests are authorized by the service."
  }
 ]
}