OpenAI's Control Room: every research inference request now needs a live job authorization, so a stop order halts model output within 35 seconds

official (OpenAI, Nov 9, 2026); company description of its own internal system, not independently verified2026-11-09

Design choice: the kill switch sits in the small, centralized inference layer, not in fast-changing workload code. OpenAI also runs sweeps for rogue engines that skip the check.
Its misalignment monitors can stop jobs automatically; OpenAI says its recent DNS-related incident was caught by a monitor.
For teams running agents: the same pattern applies at small scale. Give each agent run an ID and make the model gateway refuse requests for revoked IDs, instead of trusting the agent loop to exit.