Satya Nadella’s “emergency brake” means keeping a human-controlled way to pause or stop an advanced AI system while it is working. He proposes it because powerful systems could be compromised or behave unsafely, so people should be able to monitor and contain them rather than rely on trust alone, as reported by CNBC and The Verge. His broader safeguards include model diversity, a human-readable record of actions, continuous testing, independent controls and audits, containment, and incident disclosure. This is a safety and governance proposal; the available sources do not establish that Nadella proposed slowing AI because Microsoft miscalculated its business prospects.