Microsoft CEO Satya Nadella called on October 10, 2026 for a new trust architecture for frontier AI systems, centered on an emergency brake allowing authorized personnel to pause or shut down an AI model mid-task. He also proposed tamper-proof human-readable logs and independent audits of AI deployments. Microsoft President Brad Smith had already backed the emergency brake idea in September 2026, with similar warnings from Elon Musk and Sam Altman. The main concern centers on agentic systems and the risk of loss of control over AI that keeps acting without anyone being able to stop it. Open questions remain, including who qualifies as authorized personnel and what standards should apply to independent audits.
Source: Read the original article
