1 of 1
Story summary
- Microsoft CEO Satya Nadella proposes an architecture that lets authorized humans pause or shut down AI models mid-task.
- He likens the brake to a car’s emergency stop and demands tamper-proof, human-readable logs.
- Nadella urges independent third-party audits and assumes models are compromised from the start.
- Critics warn smaller firms may struggle with costly logging and audit requirements and question authorized personnel criteria.