Microsoft’s Satya Nadella says AI models need an ‘emergency brake’

AI agents become risky when the model that decides what to do is wired into the systems that grant permissions. Microsoft CEO Satya Nadella proposes separating the model from that surrounding software, logging every meaningful action as tamper-proof evidence a person can read, and always keeping a way to pause a model mid-task; assume a model is compromised, he wrote, and contain it like an emergency brake.

Summary written with AI by HowTechMoves, based on reporting by TechCrunch.