Artificial intelligence is the future of war? The Pentagon is putting pressure on Silicon Valley

Ashley Davis27/02/2026

0 72 4 minutes read

For several weeks, Anthropic has been opposing the demands of the American government, which wants to use the company's artificial intelligence for military purposes and mass surveillance of civilians. The company, which has become the main rival of OpenAI and ChatGPT thanks to its “Claude” model, does not intend to help either spy on citizens in the country or develop autonomous weapons systems.

CEO Dario Amodei introduced strict rules for the use of AI at an early stage: the company was born with a purpose creating particularly safe artificial intelligence. Now, however, US Secretary of Defense Pete Hegseth, who calls himself Secretary of War, is bringing out the big guns.

If the company doesn't give the Pentagon unrestricted access to its algorithms, they threaten it severe sanctions, including the application of the Defense Production Actoriginally created to secure products crucial to waging war. This is reported by the American news platform “Axios”, citing government sources.

Anthropic's back to the wall. Hegseth gives you until Friday

The dispute between the Pentagon and Anthropic is not just an abstract debate about ethics – it is taking place in the context of a contract worth up to $200 million. (approx. PLN 715 million) that the US Department of Defense signed with this company in the summer of 2025 to jointly develop AI capabilities for national security purposes.

It's about AI applications, including: for data analysis, situation assessment or automation of complex decision-making processes. According to reports, Claude is currently the only AI model that can be used in secret, isolated Pentagon systemsand military employees praise its capabilities “with only superlatives.”

The article continues below the video

It is this success that has sparked the conflict: the Department of Defense is demanding that it be allowed to use Claude “for all lawful purposes”—that is, without any restrictions resulting from the rules of use. The Pentagon recognizes that Anthropic's ethical framework is currently 'too restrictive'because the company categorically excludes the use of its AI to develop weapons that can operate without human intervention.

Anthropic explains this with its own security rules and fear of abuse. Initially, the Department of Defense threatened to terminate the framework agreement if Anthropic did not agree to the requested expanded access. But now Pete Hegseth is going to go further – either force the company to cooperate under the Act or blacklist itwhich would exclude Anthropic from further government contracts and cooperation with government suppliers.

This situation is therefore becoming a test of the relationship between the state and technology companies. Artificial intelligence is already becoming a strategic power resource — whoever controls the most advanced models gains a military advantage in the area of reconnaissance, cyber defense and coordination of operations.

For the Pentagon, what is at stake is not just one startup company, but… maintaining technological capabilities in global competition, especially towards China. The fact that cutting-edge AI is being developed almost exclusively in the private sector is shifting the balance of power – suddenly it's startup founders who decide who gets access to key solutions. Hegseth does not intend to allow this and – according to media reports – has given Anthropic an ultimatum with a deadline of Friday.

Should we cooperate or let China overtake us? The dilemma of technology companies

Anthropic is not alone in its restraint, although few companies set the boundaries so clearly. OpenAI recently relaxed its usage rules and is also working with the Department of Defense to develop voice controls for drones. emphasizing that it does not support the construction of autonomous weapons systems.

In 2018, Google withdrew from the Maven drone project, but today it is again providing cloud and AI infrastructure to the needs of government institutions. Microsoft has been working closely with the Pentagon for yearsamong others for cloud services and data analysis. Companies such as Peter Thiel's Palantir or the defense startup Anduril openly position themselves as technological partners of the army. Anduril and Microsoft are working together on AI glasses for soldiers.

The Pentagon building in Washington. Illustrative photoIvan Cholakov / Shutterstock

The close cooperation between the technology industry and the military has repeatedly been a source of internal tensions. Thousands of Google employees protested against the Maven drone project. Open letters against military contracts and orders for security services also appeared at Microsoft and Amazon.

Critics warn against a gradual normalization of the use of AI in autonomous weapons systems and state surveillance software. Supporters respond that democracies cannot place vital security technologies in the exclusive hands of authoritarian states. In China, employees of technology companies do not protest against military projects, and the cooperation of AI start-ups with the Chinese People's Liberation Army is obvious there.

Europe draws a 'red line'

Meanwhile, in Europe, the military use of AI is being considered in a different context. In an analysis published in February 2025 Scientists from the European Parliament Research Office presented a much more regulatory approach. The document highlights the EU's “people-centred” framework.

While the EU Artificial Intelligence Act does not directly apply to military applications, its ethical guidelines, according to the authors, also constitute a reference point for the defense sector. Mandatory human control over safety-related systems remains crucial.

The European Parliament takes a particularly clear position on autonomous weapons systems – has been demanding an international ban on fully autonomous, lethal weapons for years. Thus, Europe marks a “red line” that is not drawn so sharply in the American strategy. However, the question arises whether a strictly regulated model based on maximum human control can compete with authoritarian states that use AI militarily without comparable ethical and legal restrictions?

The Pentagon itself apparently answers this question in the negative. Also in Europe, start-ups such as Stark and Helsing are forming the first AI drones that independently locate targets. As the war in Ukraine shows, it is this technology that is used first – regardless of regulations.

Ashley Davis27/02/2026

0 72 4 minutes read