Satya Nadella: Assume Every AI Model Is Already Compromised

Satya Nadella says we should assume all AI models are 'compromised'

In a lengthy post on X, Microsoft CEO Satya Nadella argued that advanced AI models can no longer be treated as nested black boxes whose outputs we simply accept or reject. He called for transparent systems that can be contained, observed, and audited, with tamper-proof human-readable evidence, timely incident disclosure, and independent audits. Going further than many peers, he said every model should be assumed compromised and contained from the start, with an authorized person always able to pause or shut it down mid-task.