Anthropic has scored a major victory in its legal battle with the Pentagon , as a judge granted a temporary restraining order blocking the government’s ban on the company. The ruling is a critical development in the escalating conflict between artificial intelligence companies and military authorities over the use of AI in defense applications. The case sets a precedent for the limits of government power to control private technology companies and constitutional protections for free speech.
Judge Rita F. Lin of the Northern District of California ruled that Anthropic ’s punishment for “ hostile press behavior ” amounted to “ a classic form of retaliation First Amendment .” The ruling, which will take effect in seven days, protects the company from being banned while the lawsuit is pending. Defense Department documents revealed that the decision to designate Anthropic as a “ supply chain risk ” was based explicitly on the company’s public criticism of the government, which the judge found constitutionally unconstitutional.
See also: Anthropic vs. Pentagon: Support from Google and OpenAI
The controversy began when the Department of Defense demanded that Anthropic remove safeguards from its Claude, allowing for “lawful use of AI” and the development of fully autonomous weapons without human supervision. The company has flatly refused, considering such applications ethical red lines. The demand arose from a memorandum by Defense Secretary Pete Hegseth on January 9, calling for the inclusion of the term “any lawful use” in all contracts for the provision of AI within 180 days, including existing contracts.
The Pentagon has threatened to designate Anthropic as a supply chain risk, cancel the $200 million signed in July 2025, and invoke the Defense Production Act of 1950 — a Cold War law that allows the president to compel private companies for national defense purposes. This law has been used more than 100 times since 1950.
See also: Anthropic denies Pentagon's new terms for autonomous weapons
Anthropic resists Pentagon pressure
Anthropic itself in the AI industry with its Claude that have built-in protections against misuse. The company was the first to integrate AI models into classified military workflows, but refuses to grant unfettered access for applications such as autonomous lethal weapons and mass domestic surveillance. This stance reflects the company’s philosophy of “Constitutional AI,” which aims to create AI systems that respect human values and rights.
Anthropic CTO Dario Amodei publicly rejected the concessions, calling the Pentagon ’s assurances insufficient despite written acknowledgements of U.S. surveillance laws. In retaliation, the Pentagon CTO called Amodei a “liar.” The public confrontation was a key element of the court ’s ruling, as Judge Lin emphasized that the government cannot punish companies for expressing criticism through the media.
Impacts on the artificial intelligence sector
The defense AI sector is expanding rapidly amid the US-China technology competition , with competitors accepting the Pentagon 's terms for unrestricted access.
Anthropic is at risk of being cut from the $1.8 billion military AI budget in fiscal year 2026. The company argues that the ban will damage its reputation among ethically-minded businesses, while DoD users will temporarily lose access to Claude . 70% of DoD AI pilots now include foundational models like Claude , making the loss of access particularly problematic .
See also: Anthropic: Claude AI uses your Mac when you're away
Technical challenges and data protection
Integrating AI models into distributed networks presents complex technical security challenges. Anthropic ’s safeguards include algorithms that identify and deny requests to create content that could be used in malicious applications. Removing these safeguards would require extensive reprogramming of the model, essentially creating a new version without ethical constraints . For businesses using AI systems, it is recommended to implement multiple layers of controls, regularly evaluate the results, and maintain human oversight of critical decisions.
The temporary restraining order preserves access during the trial, protecting Anthropic ’s reputation . The pending trial is expected to test the limits of the Act on issues of software ethics and First Amendment protections . The case highlights the fundamental tension between national security and technological innovation, with implications that extend beyond the defense sector.
