This is a developing story. NCT previously covered the White House meetings with Anthropic CEO Dario Amodei on May 1 and the NSA’s testing of Mythos.

The Trump administration has formally told Anthropic it opposes the company’s plan to expand access to Mythos, its autonomous cybersecurity AI model, from approximately 50 organizations to roughly 120. The Wall Street Journal first reported the opposition, which Futurism and The Next Web confirmed through additional administration sources.

Two Grounds for Opposition

The administration cited two specific concerns, according to The Next Web. First, security: Mythos can autonomously find and exploit vulnerabilities across critical software, and broadening access increases the chance of misuse or unauthorized proliferation. The model discovered thousands of zero-day vulnerabilities in testing, succeeded on 73% of expert-level capture-the-flag tasks, and completed a 32-step simulated corporate network attack end-to-end.

Second, compute scarcity: the NSA and other government agencies currently use Mythos, and officials expressed concern that serving 120 organizations would degrade government access. Anthropic has denied the compute constraint claim but is reportedly pursuing a $900 billion fundraising round partly to secure infrastructure capacity for Mythos at scale.

The Contradiction

The formal opposition arrives at the same time the White House is developing a separate executive action to allow government agencies to work around the Pentagon’s supply-chain-risk designation of Anthropic and expand federal use of its models, according to The Next Web. The administration is simultaneously blocking private-sector expansion while trying to increase its own access.

Pentagon CTO Emil Michael told CNBC that Mythos was a “separate national security moment” from the broader supply-chain-risk dispute, attempting to decouple the two conflicts.

White House AI advisor David Sacks tweeted last month that “a growing number of people are wondering if Anthropic is the AI industry’s ‘boy who cried wolf,’” suggesting that if Mythos threats don’t materialize, the company faces a credibility crisis.

Context: An Escalating Pattern

This opposition marks the sharpest escalation since the White House chief of staff Susie Wiles and Anthropic CEO Dario Amodei met on May 1 in what both sides described as productive. The meeting, which NCT covered at the time, focused on Mythos cybersecurity implications. Two days later, the administration moved from discussion to formal objection.

The position also coincides with a reported unauthorized access incident involving Mythos, which Anthropic is investigating. The breach, though details remain limited, reinforced government anxiety about containment as the user base grows.

The Emerging Precedent

This is the first time the federal government has formally opposed a private AI company expanding access to its own model on national security grounds. The dispute establishes Mythos as a category-defining case: an AI system whose capabilities are simultaneously coveted by government and deemed too dangerous for broader distribution. Whether that stance holds as other capable models emerge will define the regulatory framework for agentic AI access for years.

An Anthropic spokesperson told the WSJ the company is having “productive conversations” about the rollout, per Futurism.