Satya Nadella says we should assume all AI models are ‘compromised’

Microsoft's CEO has stated that highly advanced AI models should be treated as potentially compromised from the start and contained with safeguards like an "emergency brake" that allows authorized people to pause or shut down a model mid-task. His recommendations for managing AI risks include building more transparent systems, timely incident disclosure, independent audits, verifiable data, and containment technologies that should be standardized. He argues against treating AI as a "set of nested black boxes" whose advice people simply accept or reject, instead calling for systems that leave behind tamper-proof human readable evidence. More advanced models would require more advanced containment technologies according to his proposal.