Anthropic in talks with Trump administration over new AI model despite Pentagon ban
Anthropic is discussing its latest AI model, Mythos, with the Trump administration even as the Pentagon has blacklisted the company from government contracts, co-founder Jack Clark said Monday at a Washington event.
The Pentagon labeled Anthropic a supply-chain risk last month after a dispute over guardrails for military use of the company's AI tools. The ban prevents the Pentagon and its contractors from using Anthropic's systems.
Clark said the contracting dispute should not overshadow the company's commitment to national security. "Our position is the government has to know about this stuff," he said at the Semafor World Economy event. "So absolutely, we're talking to them about Mythos, and we'll talk to them about the next models as well."
The specifics of those talks - which agencies are involved and what they cover - remain unclear.
What Mythos does
Anthropic announced Mythos on April 7 as its most capable model for coding and autonomous tasks. The system can write code at a high level and identify cybersecurity vulnerabilities, according to security experts.
Those same capabilities raise questions about potential misuse. The model's ability to devise ways to exploit security flaws is unprecedented among AI systems, experts said.
The legal backdrop
A federal appeals court in Washington declined last week to block the Pentagon's blacklisting of Anthropic. The decision favored the Trump administration in what has become a legal tug-of-war: another appeals court previously ruled against the Pentagon in a separate challenge by Anthropic.
For government officials evaluating AI partnerships, the Anthropic situation illustrates the tension between national security concerns and the need for agencies to understand frontier AI capabilities. More on AI for Government and Generative AI and LLM development.
Your membership also unlocks:
