AI Weapons Dispute Settled: Anthropic Defeats the Pentagon

AI Weapons Dispute Settled: Anthropic Defeats the Pentagon

28. 8. 2026
3 minutes reading · 3 views
Listen to the article
Audio version of the article
AI Weapons Dispute Settled: Anthropic Defeats the Pentagon

A U.S. federal court on Thursday evening overturned a Pentagon decision from February of this year. At the time, the Department of Defense designated Anthropic a risk to the supply chain, effectively shutting it out of military contracts. In a lengthy ruling, Judge Rita Lin said the department had acted unlawfully and without factual support. It had punished the company for publicly criticizing the military’s approach to deploying artificial intelligence on the battlefield. According to her, invoking national security does not mean an agency can do whatever it wants and retaliate against people who criticize the government. 

History of the dispute

On February 24, Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei an ultimatum. The company had until Friday, February 27, to allow unrestricted use of its models for all purposes. However, on Thursday, February 26, Anthropic announced that it would not back down. The following day, President Donald Trump ordered federal agencies to stop using the company’s products. Hegseth then designated it a supply-chain risk. It was the first time the U.S. government had publicly taken action of this kind against a domestic company under a little-known military procurement law intended to protect military systems from foreign sabotage. 

Anthropic refused to allow the military to conduct mass surveillance of the population and develop fully autonomous weapons that select targets without human decision-making. The company argues that today’s models are not reliable enough to control weapons and that domestic surveillance violates human rights. The Pentagon, by contrast, insists that private companies must not tie the military’s hands. Anthropic filed a lawsuit on March 9, arguing that the government had violated its right to freedom of speech. The company also complained that it had no way to challenge the government’s decision, which it said violated its right to due process. 

The court’s objections

Judge Lin wrote that the agencies’ evidence was weak and described Hegseth’s action as arbitrary. In her view, the department’s conduct indicated that its primary aim was to make a cautionary example of Anthropic for criticizing the government. She also pointed to a contradiction with Hegseth’s earlier threat to invoke the Defense Production Act against the company. That would instead mean that the company was indispensable to national security. The judge was also struck by the fact that the Pentagon continued seeking to work with Anthropic even after announcing the ban. In her view, none of this was consistent with a genuine concern about software sabotage or a threat to national security.

Signs of the court’s decision had emerged earlier. At a July 30 hearing, Lin said the government’s position was troubling and ran afoul of the First Amendment. She added that the department’s evidence had actually grown weaker over time. At the same hearing, Justice Department lawyers argued that artificial intelligence models are so vast and opaque that the Defense Department cannot assess them like ordinary hardware.

The government will appeal

As expected, the government will challenge the ruling. Anthropic sued the Trump administration in both San Francisco and Washington. The Defense Department based its action on two different designations that had to be heard by separate courts. Until the Washington case concludes, the company will formally remain designated as a supply-chain risk. A three-judge panel of the appeals court there has already heard arguments from both sides. Neither the Justice Department nor the Pentagon immediately commented on the situation. An Anthropic spokesperson said the company welcomed the decision and wanted to continue focusing on working with the government to ensure national security. 

Anthropic’s management says the ban could cost the company billions of dollars in lost business and damage its reputation. However, Thursday’s verdict removed a major obstacle to its planned initial public offering, which is expected to be among the largest in history. Although the company has not slowed down since being blacklisted, restoring its relationship with the Pentagon will give it access to contracts that had been out of reach for six months. Management says the lawsuits were intended to restore the business relationship to its previous state. However, it does not want to force the military to resume cooperation.

Sources: nytimes.com, theguardian.com and cnbc.com

Category:AI
Did you enjoy this article?
Discover more interesting posts on our blog
Back to blog

Related posts

Seventeen Times AI Went Rogue and Attacked Other CompaniesSeventeen Times AI Went Rogue and Attacked Other Companies
AI safety tests turned into attacks on real companies. Seventeen incidents reveal not only the models’ unexpected autonomy, but also the legal chaos surrounding the damage.
4 min read
1. 9. 2026
Anthropic Lets AI Agents Operate Machines. Model Hardware Standard Will Run Your WorkspaceAnthropic Lets AI Agents Operate Machines. Model Hardware Standard Will Run Your Workspace
Anthropic’s new standard connects AI agents to laboratory and manufacturing equipment. Instead of months of programming, it takes just hours—and an autonomous system can manage the entire setup.
8 min read
1. 9. 2026
Zuckerberg Wanted to Replace Up to 60 Percent of Staff With AI Agents. The Plan Failed at Its CoreZuckerberg Wanted to Replace Up to 60 Percent of Staff With AI Agents. The Plan Failed at Its Core
Meta wanted AI agents to handle most routine work and significantly shrink its teams. But the technology failed at key tasks, and employees pushed back against the plan.
6 min read
1. 9. 2026
Přihlaste se k odběru našeho newsletteru
Zůstaňte informováni o nejnovějších příspěvcích, exkluzivních nabídkách, a aktualizacích.
CodedTrip

Operated by CodedTrip LLC, USA.

YouTube
TikTok