Updated

Microsoft CEO Satya Nadella published a lengthy post on X arguing that advanced AI models should now be assumed to be potentially 'compromised' and that the industry needs to reassess the 'trust architecture' underpinning AI deployment. He called for moving away from treating AI as impenetrable 'black boxes' whose outputs are simply accepted or rejected, instead pushing for systems that can contain, observe, and produce tamper-proof, human-readable evidence of model behavior. Nadella's remarks, posted over the weekend, echo growing concern across the AI industry about losing control over increasingly autonomous systems, and align with safety measures other major labs have begun implementing.