AIUpdated Yesterday

OpenAI can stop research output within 35 seconds of a stop decision.

OpenAI says Control Room is already deployed across its research infrastructure, where it tracks model workloads and can stop output when safety or security concerns arise. The clock behind the 35-second figure starts later than an alert: it begins once there’s a decision to stop or pause a job. From that point, OpenAI says, the output can be halted within 35 seconds. Cointelegraph Arabic says people or automated monitoring systems can flag concerns. OpenAI describes external model-serving separately, so this timing is about internal research work.

2 publishers2 reports

Coverage started November 9