A U.S. federal court has issued a landmark ruling, overturning the Department of Defense's sanctions against the company Anthropic. Judge Rita Lin ruled that designating the AI developer as a "supply chain risk" was unjustified and violated the company's constitutional rights. All related restrictions have been immediately suspended.
The court's key argument — the Pentagon's dispute with Anthropic lay not in the realm of technical reliability of models or contract performance, but concerned the company's public stance on acceptable scenarios for the use of artificial intelligence. In essence, the agency attempted to punish the developer for its views, which directly contradicts the First Amendment to the U.S. Constitution, which guarantees freedom of speech.
Moreover, the court found procedural violations: Anthropic was not notified in advance of the claims and had no opportunity to contest them, which constitutes a violation of the Fifth Amendment regarding due process. The ruling took effect immediately, although the administration of President Donald Trump retains the right to appeal.
The Essence of the Conflict
The root of the disagreement lies in the terms of use of Claude models by the military. Anthropic allowed the use of AI in the sphere of national security but refused to lift bans on mass surveillance of American citizens and fully autonomous weapons systems. In February, the Pentagon issued an ultimatum, demanding the removal of all restrictions for "any lawful purposes," and after CEO Dario Amodei refused, it carried out the threat.
Later, Trump ordered all federal agencies to cease using Anthropic technologies within six months. The conflict shook the entire industry: OpenAI CEO Sam Altman publicly urged the military not to place the competitor on a "blacklist," and then admitted that certain aspects of his own agreement with the Pentagon were a mistake.
Notably, despite the sanctions, Anthropic's technologies continued to be used in government agencies — in April, the U.S. National Security Agency used the Mythos model. And in early August, representatives from Meta, Anthropic, Google, and OpenAI met with the Trump administration to discuss voluntary safety testing of the most powerful AI models.
My analysis: This ruling is a precedent for the entire AI industry. It clearly signals that the state cannot use administrative levers to suppress developers' positions on ethical issues. However, one should not forget that an appeal from the administration is likely, and the final word on this matter has not yet been spoken. Moreover, such conflicts highlight the growing tension between the commercial interests of AI companies and state policy in the sphere of security.