Federal Judge Rules Trump Administration's Export Control Order Against Anthropic Was Illegal and Unconstitutional

For the millions of developers and enterprise customers who rely on frontier artificial intelligence, the sudden disappearance of a state-of-the-art model is highly disruptive. When the U.S. government forced Anthropic to pull its highly anticipated Fable 5 and Mythos 5 models offline in June, it created a chilling effect across the global tech sector. The unprecedented move raised the alarming prospect that any commercial AI system could be abruptly banned under national security pretexts, fundamentally altering the risk calculus for businesses building on top of these foundation models.

That sweeping precedent has now been halted by the federal judiciary. On August 28, 2026, U.S. District Judge Rita F. Lin of the Northern District of California ruled that the Trump administration's export control directive against Anthropic was both illegal and unconstitutional. In a highly anticipated decision, the judge granted a preliminary injunction barring the administration from enforcing the global ban on the Claude models, delivering a significant legal victory to the San Francisco-based AI firm and its commercial partners.[1][2]

The court's decision hinged on the constitutional limits of executive power when applied to domestic technology companies and their internal safety policies. Judge Lin described the government's actions as “broad punitive measures” that appeared arbitrary, retaliatory, and designed to severely harm the company. Crucially, the ruling challenged the administration's authority to designate an American company as a potential adversary simply because its leadership expressed disagreement with government policy, framing the export control order as an unconstitutional penalty on speech.[1]

The underlying dispute that triggered this legal battle traces back to early 2026, when the Pentagon demanded that Anthropic waive its strict contractual restrictions on how its models could be deployed. Specifically, the Defense Department sought to use the Claude models for mass domestic surveillance of American citizens and for fully autonomous weapons systems that would operate without human oversight over targeting and firing decisions. These demands directly conflicted with the core tenets of Anthropic's safety-focused mission.[2]

Anthropic flatly refused the Pentagon's demands, citing its Constitutional AI framework—which bakes ethical refusals directly into the model's training—and its public commitments to AI safety. In response to that refusal, Defense Secretary Pete Hegseth designated Anthropic as a “supply chain risk” in March 2026, marking the first time a U.S. company had received such a severe national security label. The administration subsequently ordered federal agencies to phase out all use of Anthropic's technology within six months.[1][3]

The administration dramatically escalated the conflict on June 12, 2026, when Commerce Secretary Howard Lutnick issued an emergency export control directive. The order suspended all access to Anthropic's newly released Fable 5 and Mythos 5 models by any “foreign national.” In a highly unusual application of the law, this restriction applied not just to international users, but to non-U.S. citizens working inside Anthropic's own American offices, effectively paralyzing the company's internal operations and development pipelines.[2][5]

The specific mechanism of the export control order made compliance technically impossible without initiating a total shutdown. Because Anthropic could not reliably distinguish foreign nationals from U.S. persons in real time across a massive user base of hundreds of millions operating on shared cloud infrastructure, the company was left with only one viable option. Anthropic was forced to execute a hard global shutoff of both models for all customers, pulling them from the market just five days after launch.[2][6]

Anthropic, joined by commercial customers such as the AI litigation firm Legion, immediately challenged the directive in federal court. They argued that the administration had vastly exceeded its authority under the Export Control Reform Act of 2018, which was designed to regulate the transfer of physical munitions and sensitive hardware, not commercial software. The plaintiffs contended that the government's actions were a retaliatory abuse of power aimed at forcing compliance with military demands.[4]

The evidence presented during the hearings strongly supported the plaintiffs' claims of retaliation. Judge Lin noted in her ruling that the specific military authority used by the administration had previously only been deployed to target foreign adversaries, making its application to a domestic tech company highly irregular. The ruling explicitly questioned why Anthropic was being penalized specifically after attempting to limit its technology from being used in fully autonomous weapons or domestic surveillance programs.[1]

The government's stated rationale for the emergency export control order was that a “jailbreak” had been discovered that bypassed the models' safety guardrails, allegedly unlocking dangerous cybersecurity capabilities. However, the evidence supporting the severity of this claim appeared thin during the proceedings. Anthropic countered that the vulnerability in question was minor, already publicly known, and easily reproducible on competing models like OpenAI's GPT-5.5, which faced no such government restrictions or export controls.[6]

Further weakening the government's case was the lack of transparency surrounding the alleged threat. The government has not released its evidence regarding the specific national security concern publicly, citing classification rules. Anthropic noted in its legal filings that the Commerce Department's letter “did not provide specific details of its national security concern,” leaving the company in the impossible position of having to comply with severe restrictions that it could not fully evaluate or challenge on technical merits.[5][6]

At the center of this legal dispute is a profound constitutional question that will likely shape the future of the AI industry: whether national security powers can be used in a way that indirectly penalizes speech and corporate ethical stances. Judge Lin's ruling strongly suggests that First Amendment and due process protections extend to a company's right to enforce its own safety guardrails against government demands, setting a critical boundary on executive overreach.[1]

The case has also intensified international debates over technological sovereignty and the reliability of U.S. infrastructure. European analysts warned that the U.S. government's willingness to abruptly cut off global access to a commercial AI model demonstrated the severe risks of relying on American technology. The export control order effectively treated commercial software like munitions, raising fears of a new technological Cold War and accelerating calls for Europe to develop its own independent AI ecosystems.[6]

While Judge Lin's ruling grants a preliminary injunction that halts the export controls, the broader legal battle is far from over. The Justice Department is expected to appeal the decision to a higher court, and a separate lawsuit regarding the Pentagon's initial “supply chain risk” designation remains ongoing in the D.C. Circuit. The administration has signaled that it views operational control of frontier AI as a non-negotiable requirement for military readiness, ensuring further clashes.[1][2][6]

For now, the federal court's ruling clears a legal path for Anthropic to restore access to Fable 5 and Mythos 5. However, the exact timeline for when these state-of-the-art models will actually return to enterprise customers and developers remains uncertain. The company must navigate the complex technical logistics of reversing a global shutdown while bracing for further government appeals, leaving the AI industry in a state of cautious optimism mixed with lingering regulatory anxiety.