The White House wants to put Anthropic’s newest AI model, called Mythos, into the hands of federal workers, but a legal battle over the company’s government contract and an unexplained presidential directive have turned what should be a straightforward procurement into one of the most contested technology deals in recent federal history.
As of early April 2026, Anthropic’s existing Claude AI tools are back on the government’s digital shelves after a federal judge blocked an attempt to remove them. Yet the path forward for Mythos, a more advanced model that Anthropic has pitched directly to senior White House staff, remains undefined. No deployment timeline, no budget allocation, and no formal pilot program have been announced publicly.
The deal that started it all
The groundwork was laid in the summer of 2025. The administration released its national AI strategy in July, a document that framed rapid adoption of commercial AI tools across the executive branch as a strategic priority rather than a long-term aspiration.
The General Services Administration moved quickly to turn that ambition into a contract. On August 12, 2025, GSA announced a governmentwide agreement with Anthropic, branded as the OneGov deal, that offered Claude for Enterprise and Claude for Government to all three branches of the federal government for a nominal price of $1. Agencies could buy in through the Multiple Award Schedule, a standard GSA purchasing vehicle, without running their own lengthy acquisition processes.
For departments that had spent years navigating slow, expensive software procurements, the deal was unusually frictionless.
Mythos enters the picture
The White House then took its interest a step further. Chief of Staff Susie Wiles met directly with Anthropic CEO Dario Amodei to discuss Mythos, the company’s next-generation AI model, according to reporting by the Associated Press. A White House official told the AP that evaluation and technical review requirements would apply before any federal deployment. Anthropic issued a separate statement confirming it had shared information about the technology with the administration.
The meeting marked a shift. Procurement officers and contracting specialists typically drive federal technology deals. When the White House chief of staff sits down with a company’s CEO to discuss a specific product, the signal is that political leadership views the technology as significant enough to manage from the top.
What Mythos actually offers beyond Anthropic’s existing Claude models has not been detailed in any public government document. Neither the White House nor Anthropic has disclosed what technical benchmarks the model would need to clear, which agencies might pilot it first, or whether it represents an incremental upgrade or a fundamentally different product.
The ban and the court order
Whatever momentum the deal had built came to an abrupt stop on February 27, 2026, when GSA pulled Anthropic from USAi.gov and from the Multiple Award Schedule, according to the agency’s own announcement at the time. GSA said the removal was carried out pursuant to a presidential directive. The agency offered no further explanation of the directive’s rationale, and the directive itself has not been published.
The move cut off the procurement channel that had been open since the previous summer. Agencies that had been onboarding Claude tools were left without a valid contract vehicle.
Anthropic challenged the removal in court. On March 26, 2026, a federal judge in the Northern District of California issued a preliminary injunction ordering the government to halt the ban, according to the court’s docket in the case. Preliminary injunctions typically require a judge to find that the plaintiff has raised serious legal questions and faces potential irreparable harm, a meaningful threshold that suggests the court saw significant problems with the government’s action.
GSA complied. On April 3, 2026, the agency published a statement confirming it had withdrawn its removal announcement and restored Anthropic’s offerings for federal buyers, according to GSA’s public communications. The status quo that existed before the ban was back in place.
Open questions agencies are asking
The reinstatement resolved the immediate crisis, but it left a stack of unanswered questions that federal technology officers and contracting teams are now working through.
What triggered the ban? The presidential directive behind the February removal has not been published or explained in detail. Whether it targeted Anthropic specifically or applied more broadly to AI vendors is unclear from available government records. The underlying legal dispute appears to be ongoing, and GSA’s April statement did not indicate whether the administration plans to appeal or pursue the removal through other channels.
How does Mythos fit into the existing contract? The OneGov deal covers Claude for Enterprise and Claude for Government. Whether Mythos can be delivered under the same agreement or requires a separate procurement action is an open question. Neither GSA nor Anthropic has addressed it publicly. For agencies already using Claude, the distinction matters: an incremental upgrade might flow through existing contract terms, while a new product could trigger a fresh acquisition process with its own timeline and approvals.
Who is paying, and how much? The $1 price tag on the OneGov deal covered licensing, but the actual cost of deploying AI tools across federal agencies, including integration work, security accreditation, and staff training, has not been disclosed. No public Office of Management and Budget guidance or spending allocation for Mythos adoption has appeared in official channels. Financial news outlets have referenced internal OMB communications, but no primary government document confirming budget commitments is available.
Will agencies move at different speeds? Departments with operational missions may push to integrate large-scale AI more aggressively than purely administrative offices. But no cross-government implementation memo has clarified whether early adopters will be encouraged to experiment or held back by centralized risk controls.
What the sources actually show
The strongest evidence in this story comes from official government records. The White House AI strategy is a primary policy document. The GSA press releases, both the original OneGov announcement and the April 2026 reinstatement statement, confirm specific procurement actions and legal developments with exact dates. These represent binding government positions.
The Associated Press report on the Wiles-Amodei meeting is institutional journalism with named sources and attributed statements from both the White House and Anthropic. It confirms the meeting and its subject, though both sides hedged carefully. The White House official referenced evaluation requirements without committing to a deployment plan. Anthropic acknowledged sharing information without specifying what was shared or what commitments were made. Because no direct AP link to the specific article is available for verification, readers should note that the attribution rests on the AP’s institutional reporting rather than a linkable primary source.
References to OMB emails and internal memos that have appeared in financial coverage lack primary documentation. Until those documents surface through official channels or Freedom of Information Act requests, they should be treated as unconfirmed leads rather than established facts. Court filings from the Northern District of California case, as they become more fully available, will likely provide the clearest picture of how the administration justified its directive and how Anthropic framed its legal challenge.
Contracts restored, operational details still missing
For federal employees and contractors working with AI tools right now, the situation is straightforward on paper and murky in practice. Anthropic’s Claude products are available through GSA channels following the court-ordered restoration. Agencies can continue using existing contract vehicles.
But the policy environment around Mythos is unsettled. No implementation roadmap has been published. The courts have not resolved the challenge to the presidential directive. And the White House has signaled high-level interest in the technology without committing to a concrete plan for getting it into agency systems.
Until those pieces fall into place, the federal government’s relationship with Anthropic will sit in an unusual limbo: contracts restored, ambitions declared, and nearly every operational detail still to be determined.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
