Federal agencies could soon gain access to a powerful Anthropic AI system called “Mythos,” according to a White House memo that Bloomberg reports it has reviewed. The memo has not been independently confirmed by a second outlet, and its full text has not been released publicly. The guidance is reportedly moving forward even though cybersecurity concerns about Anthropic’s technology remain unresolved inside the government, and just weeks after a federal judge blocked the Pentagon from restricting the company on national-security grounds.
The result, if Bloomberg’s account is accurate, is a striking policy collision: one arm of the executive branch tried to shut Anthropic out of federal procurement, lost in court, and now the White House appears ready to push the company’s AI deeper into government operations than ever before.
What the memo signals
Bloomberg’s report, published in April 2026, describes a White House initiative to distribute a version of Anthropic’s Mythos model to major federal agencies. The memo has not been released publicly, and no second outlet has independently confirmed its full contents or the list of agencies in line for access. Bloomberg did not detail what technical safeguards or audit requirements the guidance includes, leaving open the question of whether the plan addresses the cybersecurity risks that triggered the original restriction.
Anthropic, the San Francisco-based AI company behind the Claude family of models, has not publicly commented on the terms under which it would accept large-scale federal deployments. The company has published a Responsible Scaling Policy that commits it to pausing deployment of models that cross certain capability thresholds without adequate safeguards, but it is unclear whether Mythos has undergone the kind of independent technical review that policy envisions.
The court fight that preceded it
The White House memo lands against the backdrop of a legal battle that reshaped the government’s posture toward Anthropic in a matter of weeks.
In early 2026, the Pentagon moved to designate Anthropic as a supply-chain risk, a classification that would have barred agencies from initiating new contracts or extending existing ones involving the company’s products. A federal judge in the Northern District of California issued a preliminary injunction blocking that designation, according to the Associated Press. The AP reported that the court found the restriction appeared punitive rather than grounded in the kind of narrow procurement vulnerability that federal law requires.
The statute at the center of the dispute, 41 U.S. Code Section 4713, defines “supply chain risk” as the risk that an adversary may sabotage, maliciously introduce unwanted function into, or otherwise subvert the design, integrity, manufacturing, production, distribution, installation, operation, or maintenance of a covered article so as to surveil, deny, disrupt, or otherwise degrade the function, use, or operation of the article or a system containing the article. The statute also requires agencies to follow procedural steps before restricting a vendor, including notice, an opportunity to respond, and internal review. The judge’s preliminary finding, as reported by the AP, suggested the Pentagon had not fully met those requirements.
Following the injunction, the General Services Administration issued a statement saying it would restore Anthropic technology to the status quo that existed before what it described as a February 27, 2026 cutoff, citing what it identified as Case No. 26-cv-01996-RFL. GSA said it would continue allowing integrations with Anthropic products across the government-wide acquisition vehicles it oversees. These details are drawn from the GSA statement as summarized in news reports; the underlying court docket entry has not been independently verified for this article.
What is still missing from the public record
Several gaps make it difficult to assess whether the White House guidance represents a responsible expansion of AI capability or an end run around unresolved security concerns.
First, there is no public description of what version of Mythos would be deployed, what workloads it would handle, or whether its use would be confined to low-risk environments like internal research and prototyping. Without those parameters, outside observers cannot gauge the sensitivity of the data the system might process.
Second, neither the Pentagon nor Anthropic has publicly detailed the specific cybersecurity vulnerabilities that prompted the original supply-chain designation. The government’s full legal theory for the restriction has not appeared in publicly available filings beyond what news outlets have summarized. Whether the administration views the court’s injunction as a temporary setback or a signal to rethink its approach remains unclear.
Third, GSA’s restoration order returned agencies to the pre-February 27 arrangement but did not announce new monitoring requirements, security benchmarks, or conditions. That raises the question of whether any lessons from the supply-chain review will shape future procurement decisions, or whether agencies will simply resume prior practices.
A separate legal proceeding in Washington, D.C., involves a different regulatory mechanism for restricting AI vendors, according to the AP. If the two cases reach conflicting conclusions about how far agencies can go in limiting vendors on cybersecurity grounds, federal procurement officials could face contradictory legal obligations.
The broader competition for federal AI contracts
The push to deploy Mythos does not exist in a vacuum. Anthropic is one of several AI companies competing for federal business alongside OpenAI, Google, and defense-focused firms like Palantir. The speed at which the White House is moving to expand Anthropic’s footprint suggests the administration views the company’s technology as a strategic priority, not just another procurement option.
That urgency, however, is running ahead of the public explanation. No independent technical audit of Mythos has been disclosed. No on-the-record statement from Anthropic has addressed the specific terms of federal deployment. And no congressional hearing or oversight report has examined whether the current legal and cybersecurity framework is adequate for the scale of AI integration the White House appears to envision.
Where this stands as of May 2026
The picture that emerges is of an administration determined to embed advanced AI into federal operations, constrained by statutory guardrails and judicial oversight, and moving faster on access than on transparency. The Pentagon tried to restrict Anthropic. A court said no. GSA reopened the door. And now the White House is preparing to walk through it.
Until the administration releases more detail on how Mythos will be secured, and until the courts resolve the open questions around supply-chain authority, the federal government is expanding its use of Anthropic AI under a legal and cybersecurity framework that has not caught up with the pace of deployment. For agencies on the receiving end of that guidance, the practical question is straightforward: adopt a powerful tool now, or wait for answers that may not arrive before the next memo does.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
