Anthropic in talks with Trump administration on new AI model, says co-founder

• The Pentagon barred Anthropic and its contractors last month after disagreements over how the military could use the company's AI technology
• Mythos, announced April 7, is described as Anthropic's 'most capable model' with advanced coding abilities that could identify cybersecurity vulnerabilities and exploit them
• A federal appeals court has upheld the Pentagon's national security blacklisting of Anthropic, though details of which government agencies are involved in current talks remain unclear

Summary:

Anthropic is engaging in discussions with the Trump administration regarding its latest frontier AI model, Mythos, despite an ongoing contract dispute with the Pentagon. The Defense Department labeled Anthropic a supply-chain risk last month and banned its use by the Pentagon and contractors following disagreements over guardrails on military applications of the company's AI tools.

Co-founder Jack Clark emphasized at the Semafor World Economy event that while a "narrow contracting dispute" exists, Anthropic remains committed to national security and believes the government should be informed about its AI developments. The company plans to continue briefing officials on Mythos and future models, though specific details about which agencies are involved remain unclear.

Mythos, announced on April 7, represents Anthropic's most advanced "agentic" AI model to date, capable of autonomous action. The system's sophisticated coding abilities have raised concerns about its potential to identify and exploit cybersecurity vulnerabilities at an unprecedented level.

Legal battles are ongoing, with a Washington, D.C. federal appeals court recently upholding the Pentagon's blacklisting of Anthropic—a victory for the Trump administration. However, another appeals court reached an opposite conclusion in a separate legal challenge filed by the company.

Market Implications: The dispute highlights growing tensions between AI companies and government agencies over control and deployment of advanced AI systems. The outcome could set precedents for future public-private partnerships in AI development and national security applications, affecting the broader AI sector's relationship with federal contractors.