The Trump administration is developing ways to bypass restrictions on Anthropic that it once called essential for national security, marking an abrupt reversal on the AI company the government moved to purge just months ago.
A draft executive action now in circulation could provide cover for agencies to adopt Anthropic's latest and most capable model, Mythos, while sidestepping an Office of Management and Budget directive that blocked federal use of the company's systems. The effort reflects a behind-the-scenes push to resolve a conflict that has become increasingly untenable as federal agencies press for access to technology they view as critical.
"Save face and bring them back in," is how one source characterized the White House strategy. Last week, White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent met with Anthropic CEO Dario Amodei in what both sides described as a productive initial discussion on government-industry cooperation.
This week the White House is hosting companies across multiple sectors to shape the potential executive action and establish best practices for deploying Mythos. Those sessions include preliminary reviews of guidance designed to reverse the earlier ban on Anthropic use across government.
The White House declined to elaborate on specific plans. "The White House continues to proactively engage across government and industry to protect our country and the American people, including by working with frontier AI labs," a spokeswoman said. "Any policy announcement will come directly from the President."
Anthropic had no comment.
The about-face
The pivot is striking. The Pentagon and White House were once unified in their hostility toward Anthropic, with both characterizing the company as a security liability. That posture changed markedly once Mythos emerged with capabilities that can both automate cyberattacks and defend against them, making the company a potential strategic asset rather than a threat.
The Pentagon has sued Anthropic and remains locked in dispute with the company, yet continues to operate some of Anthropic's older Claude model within sensitive military systems. The National Security Agency is already using Mythos. Across the federal government, agencies are signaling demand for fuller access to the technology.
Sources close to the negotiations suggest that while the Pentagon maintains its hardline stance, other stakeholders view the prolonged conflict as counterproductive and are seeking an exit strategy. It remains unclear whether the steps under consideration would resolve the Pentagon litigation or simply enable other government bodies to move forward with Anthropic partnerships.
The core dispute centers on Anthropic's refusal to grant the Pentagon what it calls "all lawful purposes" access to Claude. The company insisted on excluding use for mass domestic surveillance or development of fully autonomous weapons systems. The Pentagon deemed this refusal evidence of unreliability and issued an unprecedented supply chain risk designation in response.
Meanwhile, OpenAI and Google have both agreed to the Pentagon's "all lawful purposes" standard, though both companies maintain they honor the same restrictions Anthropic sought to impose.
The Pentagon remains bound by older terms of service in its Claude deployment and does not receive the latest model updates, a source said. Negotiations between the two sides risk becoming contentious again should they resume.
Author James Rodriguez: "The White House is gambling it can thread a needle here: rehabilitate Anthropic's standing without actually resolving the Pentagon's fundamental complaint, and hope nobody notices the contradiction."
Comments