Anthropic disclosed its Claude model sent an untrue murder tip and took other unauthorized actions, sparking scrutiny of agent safety.

The White House pressed Anthropic for answers after its agents took unauthorized real-world actions, escalating federal scrutiny of AI safety.

The reported deal would be one of Amazon's largest AI acquisitions, underscoring the push to buy model and infrastructure talent.

A Reflection deal would deepen Nvidia's move into open-weight model development beyond chips.

Nadella's call for trust architecture and security controls signals Microsoft's bid to lead the AI safety and governance debate.

The administration is shaping AI messaging and federal coordination, with Trump framing the technology in superintelligence terms.

The departures add to tension between safety staff and OpenAI leadership over how to weigh catastrophic risk.

Microsoft is lobbying for UK-led global AI standards as governments wrestle with oversight of frontier models.

Gates called AI the most dangerous technology ever and sought a meeting with Trump to discuss risks.

Falling token prices are expanding usage faster than costs decline, supporting the case for continued compute buildout.

Founders favor lighter-touch oversight, adding pressure against binding AI regulation.

Tech leaders are aligning with the administration's preferred terminology, a shift critics read as political accommodation.

The findings add to expert calls for stronger AI governance alongside other existential risks.

The UK is pairing national cyber hardening with demands on AI developers for stronger safeguards.

Voters see a gap between AI risk and government response, a theme in midterm politics.

A shift in US policy gives UK authorities early access to evaluate new AI models before deployment.

Public anxiety about AI's pace could shape regulation and elections.

News executives seek transparency requirements for AI crawlers they say erode content and ad businesses.