White House blocks Anthropic's plan to expand Mythos AI model access

9 articles · Updated · The Wall Street Journal · Apr 30

Anthropic sought to extend Mythos access to 70 more entities, raising the total to about 120 organizations.
Officials cited national security and concerns over Anthropic's computing capacity, fearing expanded access could hamper government use and increase cyberattack risks.
The decision highlights ongoing tensions between Anthropic and the administration, amid legal disputes and worries about AI models' ability to find software vulnerabilities and facilitate cyberattacks.

Will government oversight on AI create a world of approved 'haves' and innovative 'have-nots'?

How can we prevent an AI arms race as private companies build tools with nation-state level power?

As AI automates cyberattacks, how can human-led security teams possibly keep up with machine-speed threats?

What happens when a rival nation develops its own AI super-weapon without any ethical constraints?

Is it possible to truly secure an AI model that is designed to be an expert hacker?