White House internal fight stalls US AI regulation after Mythos Internal conflicts within the US federal government, primarily between the Commerce Department, intelligence agencies, and pro-industry factions, have stalled the creation of federal AI regulation. This paralysis became evident with the last-minute cancellation of an executive order designed to establish pre-release safety evaluations for AI models and the removal of testing agreement announcements from NIST’s website. The development of offensive cybersecurity capabilities by AI models like Anthropic’s Mythos has heightened concerns among national security officials, who advocate for intelligence agency involvement, while the Commerce Department pushes for a civilian testing model to maintain industry trust.

• Internal factions within the Trump administration are engaged in a “knife fight” over AI regulation, leading to paralysis in federal AI policy.
• An executive order aimed at formalizing pre-release safety evaluations for AI models was abruptly canceled by Trump, who expressed concerns about slowing American AI leadership.
• Announcements regarding pre-deployment testing agreements between NIST’s CAISI and major AI companies were removed from NIST’s website without explanation.
• National security officials want intelligence agencies to evaluate frontier AI models, especially those with offensive cybersecurity capabilities like Anthropic’s Mythos, while the Commerce Department prefers a civilian testing model.
• The absence of a coherent US AI framework contrasts with the European Union’s AI Act and creates regulatory uncertainty for companies, leading states to draft their own legislation.
• Trump’s earlier repeal of Biden’s AI executive order removed the only existing federal framework, and no replacement has materialized.
• The current policy vacuum leaves the US government unable to respond coherently to significant AI capability demonstrations. Continue reading https://thenextweb.com/news/the-white-house-is-at-war-with-itself-over-who-gets-to-regulate-ai