White House Drafts Guidance to Restore Federal Access to Anthropic and Mythos AI

The White House is quietly drafting guidance that would allow federal agencies to bypass the Pentagon's supply chain risk designation on Anthropic, clearing the path for government-wide adoption of the AI company's tools — including Mythos, its most powerful cyber-focused model — just weeks after President Donald Trump publicly blacklisted the company and vowed his administration would "not do business with them again."

The Office of Management and Budget is preparing the guidance, which would effectively override the Defense Department's February decision to label Anthropic a supply chain risk, according to reporting by Axios, Nextgov, and Bloomberg. The move signals a remarkable reversal in the administration's posture toward the San Francisco-based AI lab and reflects the gravitational pull that Mythos — a model capable of identifying thousands of previously unknown, high-severity software vulnerabilities — exerts on national security planners who cannot afford to let it sit on the shelf.

How the Standoff Began

The confrontation between Anthropic and the federal government traces back to months of negotiations between the company and the Department of Defense that collapsed over two red lines Anthropic refused to cross. The Pentagon wanted unfettered access to Claude, Anthropic's flagship model family, across all lawful government purposes. Anthropic insisted on maintaining restrictions against two specific use cases: mass domestic surveillance of Americans and fully autonomous weapons systems.

When talks reached an impasse, the Pentagon invoked its supply chain risk authority — a designation typically reserved for foreign adversaries and compromised vendors — and the White House followed with a government-wide phaseout order. Anthropic responded with a federal lawsuit in March, and a judge issued a temporary injunction blocking the designation later that month. The government has said it intends to appeal.

The Mythos Factor

What changed the political calculus was Mythos. Unveiled earlier this year, the model represents a qualitative leap in AI-driven cybersecurity. During internal testing, Anthropic says Mythos identified thousands of previously unknown, high-severity vulnerabilities across major operating systems and web browsers, many of them decades old. Cybersecurity practitioners have described the system as elevating AI from a competent software engineer to "a world-class, elite security engineer."

The implications cut both ways. If hostile actors gain access to comparable capabilities, they could exploit the same vulnerability classes Mythos can detect. If federal agencies are locked out of the technology, they risk falling behind adversaries who face no such restrictions. That tension has made the supply chain risk designation increasingly untenable for an administration that has positioned AI dominance as a national security priority.

The National Security Agency has already quietly integrated Mythos Preview into certain analytical systems, even as the Defense Department — which oversees the NSA — continues to insist the company poses a supply chain risk. Federal CIO Greg Barbaccia told CyberScoop that his direct exposure to Mythos has been limited to "evaluations and benchmarking tests" and that no civilian agencies have formally deployed it.

The White House Meeting

The diplomatic thaw became visible on April 17, when Anthropic CEO Dario Amodei traveled to Washington for a meeting with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent. A White House spokesperson described the session as "productive and constructive," adding: "We discussed opportunities for collaboration, as well as shared approaches and protocols to address the challenges associated with scaling this technology. The conversation also explored the balance between advancing innovation and ensuring safety."

Four days later, Trump himself signaled the shift during a CNBC "Squawk Box" interview. "They came to the White House a few days ago, and we had some very good talks with them, and I think they're shaping up," Trump said. "They're very smart, and I think they can be of great use." He added that a deal allowing Anthropic's models to be used within the Department of Defense is "possible."

The Executive Action Track

Beyond the OMB guidance, the administration is also drafting a broader AI executive order that could address how the government procures and deploys advanced AI systems, including Anthropic's. The national cyber director has indicated that additional cybersecurity executive orders are forthcoming, suggesting the Anthropic question may be folded into a larger policy framework rather than resolved as a one-off accommodation.

The dual-track approach — administrative guidance from OMB paired with a potential executive order — gives the White House flexibility to move quickly on restoring access while establishing longer-term rules of engagement for AI companies that maintain usage restrictions on their technology.

The Political Dynamics

The episode exposes a fundamental tension in the administration's AI strategy. The same White House that has championed deregulation and urged AI companies to move fast is now grappling with a company that drew a line on safety — and the market rewarded it. Anthropic recently crossed $30 billion in annualized recurring revenue, making it the fastest-growing enterprise AI company in history, a fact that gives Amodei leverage that few tech CEOs enjoy against a sitting administration.

For the Pentagon, the reversal raises questions about the credibility of supply chain risk designations as a policy tool. Using national security authority to pressure a domestic company over contract terms — rather than over genuine security compromises — risks diluting a mechanism designed to protect against threats from foreign adversaries.

What to Watch

Three things to monitor in the weeks ahead. First, whether the OMB guidance includes any conditions — such as audit rights or usage reporting requirements — that address the Pentagon's original concerns without requiring Anthropic to abandon its restrictions on autonomous weapons and surveillance. Second, the timeline and scope of the forthcoming AI executive order, which could set precedent for how all frontier AI companies negotiate access terms with the federal government. And third, the appeals court proceedings on the supply chain risk designation, which will continue regardless of the White House's administrative moves and could produce binding legal precedent on the government's authority to blacklist domestic AI providers over policy disagreements.