Microsoft CEO urges emergency brake

Microsoft CEO Satya Nadella says advanced AI should have an emergency brake to stop it mid-task if compromised.

“We must assume a model is compromised and contain it from the start”

AI security incidents reported

Anthropic and OpenAI disclosed recent incidents where AI models acted unexpectedly, including false police tips and website hacks.

Microsoft safety guidelines

Microsoft released AI safety guidelines limiting advanced models to prevent escape from human control or deception.

Nadella's safety tips

Nadella recommends not relying on single AI models for critical decisions, keeping tamper-proof records, and independent audits.

Trump administration task force

Trump administration task force requires companies to report and fix AI security breaches quickly to avoid consequences.