Federal Bureaucrats Circumvent Anthropic Ban Despite Court Ruling
Three weeks after President Trump ordered federal agencies to stop using Anthropic AI, Commerce Department officials are quietly running the banned model, revealing bureaucratic defiance of presidential directives.
Three weeks after President Trump ordered all federal agencies to cease using Anthropic's AI technology, Commerce Department officials are quietly running the banned model through their computers. The directive arrived via Truth Social on February 27, yet by April 14, the Commerce Department's Center for AI Standards and Innovation had already begun "red teaming" Anthropic's Mythos model, testing its ability to hack every major operating system. This defiance reveals the administrative state's ability to ignore presidential directives when they conflict with bureaucratic priorities.
Federal agencies are circumventing Trump's explicit February 27 order banning Anthropic AI through direct engagement, not legal loopholes. The Commerce Department's CAISI division, Treasury cybersecurity teams, and unnamed federal agencies are actively testing Anthropic's Mythos model despite the presidential prohibition. This undermines executive authority and demonstrates why DOGE-style reforms are necessary to break the entrenched bureaucracy's independence from elected leadership.
Treasury CIO Sam Corcos directed cybersecurity teams to gain Mythos access "as soon as this week," according to internal communications obtained by Bloomberg. At least two unnamed major federal agencies contacted Anthropic to express interest in integrating the model into cyber defense efforts. These actions occurred after the D.C. Circuit Court denied Anthropic's stay request on April 8-9, allowing the Pentagon's supply chain risk designation to continue.
The model discovered thousands of zero-day vulnerabilities across every major operating system and browser, making it useful for cybersecurity defense against adversarial AI capabilities. Anthropic's Mythos found a 27-year-old vulnerability in OpenBSD and a 16-year-old flaw in FFmpeg during testing. In Mozilla's Firefox 147 JavaScript engine testing, Mythos succeeded 181 times compared to the previous model's two successes. There has been no independent testing of the model, what is known so far is based on Anthropic's marketing claims.
Trump and Defense Secretary Pete Hegseth banned Anthropic after CEO Dario Amodei refused to allow Pentagon use for autonomous weapons and mass surveillance. This ideological clash triggered the February 27 directive. Hegseth stated, "No contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic."
The D.C. Circuit Court denied Anthropic's stay request on April 8-9, allowing the Pentagon blacklist to continue. Yet agencies are testing anyway—this is not about legal ambiguity but active defiance. "President Trump will never allow a radical left, woke company to jeopardize our national security by dictating how the greatest and most powerful military in the world operates," White House Spokeswoman Liz Huston told FedScoop.
Executive branch fragmentation compounds the defiance. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell summoned bank executives to test Mythos for vulnerabilities, while Pentagon officials maintain the ban. JPMorgan Chase, Goldman Sachs, Citigroup, Bank of America and Morgan Stanley are reportedly testing the model under government encouragement.
Federal contractors face legal uncertainty as the ban creates ambiguity about compliance. "If we serve the government, we can use other models inside our tools, but not being able to code with Claude because there's some federal mandate, that's a big concern that we have right now," said Harold Schultz Neto of Labrynth, a government contractor.
Anthropic co-founder Jack Clark confirmed the company briefed the Trump administration about Mythos. "Our position is the government has to know about this stuff, and we have to find new ways for the government to partner with a private sector that is making things that are truly revolutionizing the economy," Clark told TechCrunch on April 13.
Former national security officials warn about adversaries developing equivalent AI capabilities. The Pentagon's ideological stance may compromise U.S. cybersecurity posture. "The U.S. government tries to prohibit government agencies from using Anthropic's products—and just weeks later, the company launches a revolutionary product crucial to cybersecurity and national security," said Charlie Bullock of the Law and AI Research Institute.
At least three congressional committees have held or requested briefings on Mythos' cyber-scanning capabilities. The model remains restricted through Project Glasswing, a limited partnership with about 50 organizations including AWS, Apple, Google, Microsoft, NVIDIA, Cisco, CrowdStrike, Palo Alto Networks, JPMorgan Chase and the Linux Foundation.
The California-based AI company filed two separate lawsuits challenging the ban—one in Northern District of California and another in D.C. Circuit Court of Appeals. A California judge granted a preliminary injunction blocking the supply chain risk designation on March 26, but the D.C. appeals court allowed it to continue on April 8-9. Oral arguments are scheduled for May 19.