Hegseth and White House allies renew criticism of Anthropic after model access changes

Defense Secretary Pete Hegseth renewed criticism of Anthropic on Saturday, days after the AI company said it discontinued access to two of its models to comply with a directive from the Trump administration. Hegseth posted on social media that “@DeptofWar kicked @AnthropicAI out of our building - forever,” adding that each additional day “proves why that was the right move,” according to The Hill.

The dispute centers on Anthropic’s reported decision to limit access to two models. The Hill reported that the change was made to comply with what it described as a directive from the Trump administration, and that the administration’s allies have continued to characterize the move as a validation of the government’s earlier decision to cut off Anthropic access on defense premises.

Hegseth’s comment followed the broader push by senior administration officials to tighten how outside technology providers interact with the federal government. In the same report, the defense secretary presented Anthropic’s model-access changes as evidence for the government’s stance, and suggested the removal from the building remains permanent.

The Hill further reported that Hegseth’s remarks were part of a renewed campaign of criticism by White House allies toward Anthropic. The reporting did not identify specific legal filings or agency actions accompanying the public statements, nor did it provide the text of the directive referenced in connection with Anthropic’s model-access change.

For its part, Anthropic’s reported action suggests the company complied with a federal instruction requiring it to alter how it offers access to particular models. The practical effect, as described by The Hill, is reduced availability of the two models, at least in the channels connected to the government instruction, rather than a court-ordered remedy or a legislative ban.

It was not immediately clear from the report whether the directive was issued by a specific agency under a particular statutory authority, or how long the restrictions would remain in place. The episode appears to be a public, escalating dispute between the administration and a private AI provider over what types of model access can be used in defense-linked settings.

As the public back-and-forth continues, the next steps, absent additional official documentation, will likely depend on whether the administration or the Department of Defense issues further guidance or formal policy on AI vendor access and whether the affected vendor details the compliance requirements it faced.

Still, the exchange underscores how quickly the administration’s technology governance priorities can translate into provider behavior, particularly where defense-related access and facility entry are at issue. For contractors and technology companies, changes in model access terms can carry operational consequences, including potential disruptions to pilots, integrations, and internal systems reliant on particular model capabilities.