Search Write for us Sign in
OpenAI Daily News

Independent news about OpenAI and ChatGPT. Not an OpenAI publication.

Sunday, November 15, 2026Sources checked

Models & Research

OpenAI adds a 35-second emergency brake for its research AI

CryptoSlate

OpenAI adds a 35-second emergency brake for its research AI

Picture: CryptoSlate

OpenAI says its newly disclosed Control Room system can cut off further model output from research jobs within 35 seconds after the service receives a stop instruction, placing enforcement outside the job's own code. The company says the system is already deployed across its research infrastructure. The 35-second maximum begins when Control Room receives the instruction; spotting a problem, investigating it and deciding to intervene happen before that clock starts. The design centers on a job ID. A research workload attaches that identifier when it requests model output. Before fulfilling the request, the inference engine, the software running the model, checks whether Control Room still authorizes that job. Once the job is marked as stopped, the engine denies further output for it.

This report was first published by CryptoSlate. What you see here is a brief excerpt. The full text is theirs.

Continue reading at CryptoSlate

Pass it on X Facebook WhatsApp

More reports on this

Latest in Models & Research

  1. Our new blog for science

Everything in Models & Research

API & Codex

  1. OpenAI Python SDK v3.51.0

  2. Codex CLI 0.169.0

  3. Codex CLI 0.168.0

  4. OpenAI Python SDK v3.50.0

Business