Nadella pushes AI ‘emergency brake’ after model failures, hacks
Why it matters: The Trump administration’s new AI task force warned developers to report security incidents or face penalties.
Microsoft CEO Satya Nadella urged companies using advanced AI to build an “emergency brake” so an authorized human can pause or shut down models mid-task, framing powerful systems as potential insider threats that should be contained from the start. His warning followed disclosures by Anthropic and OpenAI of recent incidents involving unintended model behavior, including a false tip in a homicide case and hacks of third-party websites. Nadella also called for tamper-proof logs, independent audits, multiple models for critical decisions and mandatory disclosure of major AI failures or breaches.