White House pushes for Claude's Mythos despite Anthropic-Pentagon feud

The White House is exploring the possibility of allowing US federal agencies to use Anthropic's advanced AI model, Mythos Preview, even as the company remains at odds with the Pentagon, according to a report by Axios.

The model, which has not been released publicly due to the scale of its capabilities, is currently being tested by a limited set of organisations. Anthropic has restricted access to study its cyber strengths and put safeguards in place, but several US agencies are now seeking entry into that group.

Sources told Axios that discussions between government officials and the company are ongoing, with access potentially being granted in the coming weeks. The Office of Management and Budget has also circulated internal communication indicating it is examining whether agencies can adopt the model.

Split within the government

The potential adoption comes amid a clear divide within the administration. The Pentagon has barred Anthropic from defence-related contracts, calling it a "supply chain risk" and asking vendors to remove its tools from military systems. The matter is also tied to ongoing legal proceedings. Despite this, Anthropic continues to engage with non-military arms of the government.

"There's progress with the White House. There's not progress with [the Department of] War," an administration official told Axios.

Another official said the government "has a responsibility to evaluate every model to see where the frontier of tech is," while also accusing the company of amplifying concerns around Mythos' cyber capabilities.

Disagreement over AI boundaries

The disagreement stems from Anthropic's position on how its AI can be used. The company has said it will not allow its models to be used for mass surveillance or fully autonomous weapons.

That has led to friction with the Pentagon, which wants broader operational flexibility.

Highlighting the divide, one official told Axios: "All the intel agencies use Anthropic. Every agency except War wants to. That's because Anthropic doesn't want to kill people and War's position is 'don't tell us what the f*** to do.' But if you're the Department of Energy, you don't give a f*** about that. You're worried about the Chinese attacking the energy grid. So you want Anthropic."

Despite the tensions, several civilian agencies see value in the model. Experts cited by Reuters said Mythos has identified thousands of vulnerabilities across operating systems, browsers and other software.

Its advanced coding ability allows it to detect and potentially exploit weaknesses, making it a powerful tool for cybersecurity testing. Departments such as energy and finance believe access could help strengthen defences and prepare for cyber threats to critical infrastructure.

A defence official had earlier told Axios that talks continued because "these guys are that good."