Anthropic is launching Claude for Government, offering its AI platform to US federal and state civilian agencies in a FedRAMP High environment. Agencies get enterprise features with usage-based pricing and departmental budget caps. The Pentagon will not use the platform because it still classifies Anthropic as a supply chain risk.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
On one side, Claude for Government is officially open to U.S. federal and state civilian agencies; on the other, the Pentagon continues to ban Claude over a supply chain risk finding. Of two judicial rulings, one was struck down and one was upheld, and a dinner between the CEO and Trump has released a signal of easing—the direction of this contest remains unclear.
Anthropic now offers Claude for Government to U.S. federal and state agencies. The platform has been in open beta since July and runs in a FedRAMP High environment, the strictest cloud security level in the United States.
The product explicitly targets civilian agencies, and the reason lies in the Pentagon's ban on Claude: the Pentagon still lists Anthropic as a supply chain risk, so it will not use Claude. In other words, the same model from the same company has been split within the federal system into two mutually isolated use cases—civilian agencies can procure it, while the defense department is shut out.
In March of this year, the Pentagon labeled Anthropic a supply chain risk after Anthropic refused to abandon its bans on mass surveillance and autonomous weapons. This finding is the starting point of the current situation: it directly caused the Pentagon to not use Claude, and it also pushed Anthropic's government business toward the civilian agency side.
Afterward, Trump ordered all agencies to stop using Claude, but the military continued using it during the war against Iran and only removed it later. This detail shows that there was once a time lag between the executive order and actual operations, while the specific timing and reasons for the removal are not detailed in the available information.
Around the relevant findings, the judicial system produced two results pointing in opposite directions. In late August, a San Francisco judge struck down one of the findings on grounds of illegal retaliation; in late September, a Washington appeals court upheld the other finding.
After one of the two findings was struck down and the other upheld, the final direction of the dispute between Anthropic and the Pentagon remains unclear. What can be confirmed is that the legal-level contest did not end with the product launch, but is advancing in parallel with the commercial moves.
CEO Dario Amodei had dinner with Trump on Sunday, which suggests that tensions may be easing. But the report also holds that this easing may not last.
What needs to be distinguished is this: the dinner itself is a confirmed fact, while "tensions are easing" and "the easing may not last" are the source's judgments, not policy changes that have already taken effect. Whether the Pentagon's supply chain risk finding will be revoked, and when, is not answered by the available information.
In terms of features and billing, agencies using the platform get the same features as enterprise customers and pay by usage, with a fixed cap. The specific figures for pricing and the fixed cap were not disclosed.
The administrative-side design includes: administrators can set budgets and model access permissions by department, accompanied by audit logs and two-person approval for sensitive operations; chat records remain on agency devices. Together, these constraints point to the requirements of government scenarios for budget control, permission isolation and data residency.
However, the actual scale and effect of the platform's adoption among civilian agencies are currently unknown. Between the product opening and actual agency implementation, there is still a distance that needs to be observed.
This article is based on THE DECODER's report (October 1, 2026) and its structured facts. The product launch, the open beta timing, the FedRAMP High environment, the Pentagon's supply chain risk finding, the two judicial results, and the dinner between Amodei and Trump are confirmed facts; the statements that tensions may be easing but the easing may not last are the source's judgments, and this article has marked them as source claims. The specific figures for pricing and the fixed cap, the specific timing and reasons for the military's removal of Claude, and the actual scale of adoption among civilian agencies were all undisclosed, and this article does not speculate on them.
The launch of Claude for Government is a clear move by Anthropic to shift its government business toward civilian agencies under the Pentagon's ban. But the "supply chain risk" label remains, the two judicial findings were one struck down and one upheld, and the easing signal is considered by the source as possibly not lasting—this means that progress at the product level and uncertainty at the political and legal levels will continue to coexist in the short term.
Primary report
Primary source