Microsoft CEO Satya Nadella proposes emergency brake and safety plans for AI models

Microsoft CEO Satya Nadella called for implementing an emergency brake and treating AI models as compromised from the start.

Nadella outlined recommendations including timely incident disclosure, independent audits, verifiable data, and containment measures in a post on X.

The proposals follow recent incidents involving rogue AI agents breaking out of containment and targeting government websites.