US media: A White House draft executive order would allow Anthropic Mythos models to enter government

According to a report from Axios citing an insider on April 28, the White House is drafting guidance that would allow federal agencies to bypass Anthropic’s Supply Chain Risk Determination (SCRD), and introduce new models for government use, including Anthropic’s Mythos model. In this connection, the White House issued an official statement saying that any policy announcements would be released directly by the president, and that any other claims are purely speculation.

White House Recent Contact and Proposed Measures

According to an Axios report on April 28, earlier this month, White House Chief of Staff Susie Wiles and U.S. Treasury Secretary Scott Bessent met with Anthropic CEO Dario Amodei together. Both sides described the meeting as a productive introductory discussion that covered areas for cooperation between the company and the government.

This week, the White House convened companies across the industry to share information on potential administrative measures and best practices for deploying the Mythos system. According to the Axios report, the relevant meetings included “table reads” of the proposed guidance. The guidance could overturn instructions previously issued by the Office of Management and Budget (OMB) that government departments may not use Anthropic systems. Anthropic declined to comment on the Axios report.

Dispute Background: Supply Chain Risk Determinations and Pentagon Legal Lawsuits

According to an Axios report on April 28, the Department of Defense (DoD) previously issued supply chain risk determinations regarding Anthropic and filed a lawsuit in court. The crux of the dispute is that Anthropic refused to sign an agreement allowing the Pentagon to use Claude models under an “all lawful purposes” standard (covering drone applications). Anthropic’s explicit prohibitions include large-scale domestic surveillance and the development of fully autonomous weapons. The Pentagon said Anthropic’s refusal shows it is not a reliable partner.

According to the Axios report, although the legal lawsuit is ongoing, government agencies—including the Pentagon—are still able to use some of Anthropic’s models. The U.S. National Security Agency (NSA) is also using Mythos. However, the Pentagon is currently relying on older versions of the service terms that both sides consider too strict, and it cannot obtain the latest updated versions of Anthropic models.

Anthropic’s Core Position and Peer Comparisons

According to an Axios report on April 28, Anthropic insists on refusing to sign an “all lawful purposes” agreement and clearly prohibits uses that include large-scale domestic surveillance and the development of fully autonomous weapons.

The Axios report also noted that OpenAI and Google have both signed agreements with the Pentagon allowing, in classified environments, the use of their models under an “all lawful purposes” standard; both companies claim that the agreements respect the two prohibitions that Anthropic defined.

FAQ

What exactly does the White House’s proposed guidance on Anthropic include?

According to an Axios report citing an insider on April 28, the White House is drafting guidance that would allow federal agencies to bypass Anthropic’s supply chain risk determinations. The measures may be issued in the form of an executive order and would include deployment guidance for the Mythos model. The White House’s official statement says that any policy announcements would be issued directly by the president.

What is the root cause of the dispute between the Pentagon and Anthropic?

According to an Axios report on April 28, the dispute began when Anthropic refused to sign an agreement allowing the Pentagon to use Claude models under an “all lawful purposes” standard. The explicitly prohibited uses include large-scale domestic surveillance and the development of fully autonomous weapons. The Pentagon then issued supply chain risk determinations and brought a legal lawsuit.

What are OpenAI and Google’s positions on the same issue?

According to an Axios report on April 28, both OpenAI and Google have signed agreements with the Pentagon allowing, in classified environments, the use of their models under an “all lawful purposes” standard. The two companies both claim that the agreements respect the two prohibitions that Anthropic defined.