Anthropic model details leaked; government sanctions partly blocked
Internal Anthropic documents describing a new AI model called Mythos were leaked in March 2026, revealing technical details and reportedly indicating significant advances. The same month, the Trump administration and Pentagon sanctioned Anthropic after the company refused to remove restrictions on military use, but a federal court ordered the sanctions temporarily blocked pending further proceedings.
What happened
Internal Anthropic documents detailing a new AI model called Mythos were leaked, containing technical details and suggesting significant capability advances.
In late February, the U.S. Department of Defense demanded that Anthropic broaden Claude's permitted uses for military applications, or face action under the Defense Production Act or designation as a supply chain risk. Anthropic declined, reiterating two policy red lines: no use for mass domestic surveillance and no autonomous weapon targeting or attack decisions without human intervention.
On February 28, President Trump issued an executive order directing all federal agencies to cease using Anthropic technology. In early March, the Defense Department formally designated Anthropic as a supply chain risk, effectively excluding it from government procurement.
Reported impact
Anthropic reported that the government's actions caused difficult-to-reverse harm to the company, its customers, and partners.
Market analysis noted that the cybersecurity sector declined in response to the leaked model details and government sanctions.
Index analysis
The leaks of Anthropic's internal documents and model details exposed technical specifications of the company's AI systems.
The designation of Anthropic as a supply chain risk raised questions about whether the executive branch could unilaterally restrict market access to AI technologies without congressional authorization or established procedures.
Governance controls
On March 26, a U.S. Federal Judge in the Northern District of California (Judge Rita Lin) granted Anthropic's preliminary injunction, blocking the federal ban on use of Anthropic technology and suspending the Pentagon's supply chain risk designation.
The court found Anthropic's arguments that the government action likely violated the First Amendment and constituted retaliatory action against the company's public policy statements on AI ethics to have substantial merit.
The injunction halts the government restrictions for seven days, allowing time for an emergency appeal.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.