Is Anthropic thawing its relationship with the Trump administration? Treasury Secretary and White House Chief of Staff personally meet with CEO Dario Amodei

U.S. Treasury Secretary Scott Bessent and White House Chief of Staff Susie Wiles personally met with Anthropic CEO Dario Amodei on the 17th, with the White House describing the talks as “constructive and productive.”
(Background: Anthropic CEO Dario Amodei: In the next 6-12 months, China’s open-source AI models will catch up to Mythos)
(Additional background: Previously faced collective employee resistance: Google returning to the Pentagon AI project: Confidential Gemini deployment negotiations underway)

At the start of the year, the Pentagon labeled Anthropic as a supply-chain risk, but the White House opened another door at the same time. The attitude of U.S. government agencies toward Silicon Valley’s top AI companies has never been more divided.

On the 17th (U.S. Eastern Time), U.S. Treasury Secretary Bessent and White House Chief of Staff Susie Wiles met with Anthropic CEO Dario Amodei at the White House. A White House spokesperson characterized the meeting as a “preliminary discussion,” describing the process as “constructive and productive.”

The root of the tense relationship: the weapons red line

The rift between Anthropic and the Trump administration began with a breakdown in a military contract negotiation. During the negotiations, Anthropic insisted on setting clear usage restrictions, refusing to allow its technology to be applied to autonomous weapons systems and large-scale domestic surveillance. This line could not be accepted by the Department of Defense.

After the negotiations fell apart, the Pentagon immediately designated Anthropic as a supply-chain risk, effectively attaching a warning label within the federal procurement system. Anthropic chose to take the fight head-on and is currently suing in court over this designated matter, challenging the Pentagon’s administrative decision.

A bridge to the silver industry: the unexpected breakthrough of the Mythos model

Another entry point for thawing relations may be in the field of financial regulation. It is understood that Bessent and the Federal Reserve Chair Powell are actively encouraging major banks to test Anthropic’s newly released Mythos model.

This move is significant. If the federal government’s financial regulatory bodies proactively vouch for Anthropic’s corporate clients, it would amount to clearing compliance-related concerns for Mythos deployment on Wall Street. Compared with the deadlock in the defense sector, this route through the Treasury Department instead becomes a side door for Anthropic to enter the federal ecosystem.

During this period, Anthropic co-founder Jack Clark has also continued to appear in government settings, showing that the company is intentionally pushing forward in multiple political tracks at the same time.

Besides the Department of Defense, every department wants to use Claude

Behind this standoff is deep disagreement within Washington. According to people familiar with the negotiations, besides the Department of Defense, “almost every department” of the federal government is interested in adopting Anthropic’s technology.

The Treasury, the Department of Justice, the Department of Energy, and other agencies have strong demand for the enterprise version of Claude, but the Pentagon’s supply-chain risk label has left some agencies’ procurement teams in a legal dilemma.

This kind of split at the institutional level reflects internal tension within the Trump administration’s AI policy: security hawks want to pressure every AI company that is unwilling to fully comply with military needs, but the pragmatic demands of other departments are quietly pushing policy in another direction.