A few weeks ago, we posted about the clash between Anthropic and the Department of War (DoW). Our assessment was premature; developments have been racing ahead.
We asserted that Anthropic’s rejection of the DoW contract over the issues of domestic mass surveillance and the deployment of autonomous weapons, and its consequent replacement by OpenAI, occurred within the context of fierce competition between AI companies. Since then, Anthropic has sued the DoW and won in a San Francisco federal court, momentarily blocking the Pentagon from labeling the company a supply-chain risk. However, the company then lost on appeal in the U.S. Court of Appeals in Washington, D.C. Anthropic’s oral argument is scheduled for May 19 in Washington, D.C., at the appeals court.
It’s puzzling to see Anthropic taking on the DoW, but even more extraordinary is that when Anthropic sued the Pentagon, its rivals – Google, Microsoft, Amazon, and Palantir – banded together to support it. Microsoft even filed an amicus curiae (“friend of the court”) brief, as did employees from Google and OpenAI; and neither company’s management objected.[1] These corporations are locked in competition with each other in global markets, all the while trying to cozy up to the Trump administration. What’s happening? Is Anthropic acting as a bellwether, setting new ethical standards for the tech industry? Why are these rivals suddenly uniting against the DoW? This exceptional clash within the core of today’s military-industrial complex requires some additional exploration.
Until its widely publicized dispute with the Pentagon, Anthropic was relatively unknown to the general public. This was because Anthropic’s strategy has been to target the rapidly growing AI enterprise market: nearly one in three US businesses paid for Anthropic’s tools in March 2026,[2] under a widely publicizedguarantee of security and trust – essential to winning their business (only recently has it springboarded into the consumer market as well). Thus, Anthropic generally flew under the public radar. In contrast, OpenAI’s strategy has focused overwhelmingly on the consumer market, in which it claims 900 million users[3] (however, as its competition with Anthropic intensified, OpenAI also began to cultivate the enterprise market).
A major plank in Anthropic’s “B2B” enterprise approach has been to leverage ties with the major tech companies, trying to get them to integrate Anthropic’s Claude AI model into their other applications.
Microsoft – which initially invested and exclusively used OpenAI – has more recently incorporated Claude into its enterprise suite. This was a boon to both companies. Microsoft is one of the largest government contractors, and its technologies are deployed across all levels of government, including the DoW and other federal agencies. If Anthropic is designated a supply chain risk and government agencies are forbidden to use it, then, Microsoft will need to find a Claude replacement and will be negatively impacted financially and strategically. Microsoft isn’t the only tech company that has incorporated Anthropic’s AI models. Others too are acting as both financial backers and users of Claude.
Google had invested more than $3 billion as of 2025 and owns a 14% stake in Anthropic.[4] Google and Anthropic have since greatly expanded their partnership, signing a deal in which Google may invest up to $40 billion, including $10 billion right away. This will further grant Anthropic access to much-needed computing power and infrastructure, including the use of Google’s Tensor Processing Units (TPUs) as an alternative to Nvidia GPUs (thus challenging Nvidia’s AI chip dominance).[5] Amazon, for its part, initially invested $8 billion.[6] Along with Google, Amazon is one of Anthropic’s cloud providers and its Claude models are used by AWS clients. The two companies then closed another circular deal, through which Amazon augmented its investment in Anthropic while Anthropic committed to spend $100 billion on Amazon’s computing power and chips.[7]
Palantir – the software provider that facilitates the tracking of immigrants for the U.S. Immigration and Customs Enforcement (ICE) agency, maintains close ties with the Israeli government, and provides technology (along with Google, Microsoft and Amazon) for Israel’s genocide in Gaza and mass surveillance of Palestinians – also backed Anthropic against the War Department.[8] Palantir itself had developed an AI-enabled military platform, the Maven Smart System (MSS), for the DoW which incorporated Claude AI. Maven is used to analyze battlefield data and to identify targets.[9]
Given these entanglements, it shouldn’t be a surprise that its interlocked rivals chose to back Anthropic: it was a matter of joint self-interest. When their profits and strategies came under threat – and not for any moral reason – the group set aside competition and challenged the government together. Underlying competitive dynamics nevertheless also persisted: this was an ad hoc arrangement aimed at protecting momentary common interests.
Anthropic’s AI model was, moreover, already deeply embedded in the U.S. military, as the company itself has affirmed: “Anthropic has much more in common with the Department of War than we have differences.”[10] So why did the company take the Pentagon to court in the first place? Was it because of pressure from some of its employees, whose pacifistic views may have matched those of many Google and OpenAI workers? Was it a function of Anthropic’s brand management, which emphasized safety and trust for its business clients – and ostensible sensitivity to morality more generally?
The answers to these questions remain open. However, Enrique Dans, a Senior Fellow with the Tech Policy Program at the Center for European Policy Analysis (CEPA), underlines that a different factor is also involved. Anthropic’s fight against the Pentagon is additionally about “control of governance.” Dans clarifies this:[11]
“LLM providers are not selling neutral infrastructure. They’re selling models with built-in constraints, policies that can change, and enforcement mechanisms that can tighten overnight. … Anthropic’s stance wasn’t simply “ethical positioning.” It was product governance. The Pentagon’s stance wasn’t simply “buyer pressure.” It was demanding control of governance.“
In other words, underlying the rhetoric about “national security” and “supply chain risks” is a fight over control: whose strategy, whose decisions, will shape AI – the War Department’s or Anthropic’s? This could be yet another reason why top AI companies are backing Anthropic: the outcome of Anthropic’s clash with the DoW could set a precedent with sector-wide ramifications that they would like to ensure turn out favorably.
Anthropic’s release of its fearsomely powerful new model raises the stakes even further. Anthropic initially restricted testing of Claude Mythos to banks and a select group of tech partners, so that they could try to secure in advance thousands of severe vulnerabilities that Mythos had detected in every major operating system and web browser.[12] Even before Claude Mythos’ limited release, Vice President JD Vance and Treasury Secretary Scott Bessent talked to major tech CEOs to discuss security issues around the AI model. Anthropic’s CEO Dario Amodei was one of participants along with xAI’s Elon Musk, Google’s Sundar Pichai, OpenAI’s Sam Altman, Microsoft’s Satya Nadella, CrowdStrike’s George Kurtz and Palo Alto Networks’ Nikesh Arora.
Anthropic responded about the meeting and said, “Bringing government into the loop early – on what the model can do, where the risks are, and how we’re managing them – was a priority from the start.”[13] Shortly thereafter, Amodei separately met with White House staff chief of staff Susie Wiles, Treasury Secretary Scott Bessent and National Cyber Director Sean Cairncross.[14] Anthropic was again publicizing its “good citizen” approach to AI.
As we’ve stressed, Anthropic has tried to distinguish itself from other AI companies by recurrently emphasizing trust and safety. However, Anthropic is not the first company to walk this tightwire. Google once used the slogan “Don’t be evil,” but it dropped the motto in 2018 to move unencumbered into every territory of profit including military markets. Anthropic has not reached its Rubicon, but perhaps it is getting close: In late February, 2026, amid its dispute with the DoW, Anthropic updated its original safety policy – called the Responsible Scaling Policy – and removed its promise to pause development of any AI model deemed potentially dangerous.[15]
There is, however, a deeper issue – one that follows from the fight over governance mentioned above.
The clash between Anthropic and the Pentagon has raised the question of whether a private company or the DoW will superintend governance of AI. However, this phrasing begs the real question – which is: Who should decide how, if, for what purposes, and for whom AI is used in the first place?
Shinjoung Yeo & Dan Schiller
[1] “Microsoft backs Anthropic in amicus brief to halt US DOD’s ‘supply-chain risk’ designation,” Reuters, March 10, 2026.
[2] Clara Murray, “Anthropic closes in on OpenAI as US business use surges,” Financial Times, 10 April 2026.
[3] Ibid.
[4] Cade Metz, Nico Grant and David McCabe, Inside Google’s Investment in the A.I. Start-Up Anthropic, New York Times, March 11, 2025,
[5] Cristina Criddle and Ryan McMorrow, “Google to invest up to $40bn in Anthropic,” Financial Times, April, 24 April,
[6] Amazon Staff, “Amazon and Anthropic deepen strategic collaboration,” November 11, 2024.
[7] George Hammond and Rafe Rosner-Uddin, “Anthropic and Amazon agree $100bn AI infrastructure deal,” Financial Times, April 21, 2026.
[8] Mike Isaac, “Silicon Valley Musters Behind-the-Scenes Support for Anthropic,’ New York Times, March 18, 2026; Marwa Fatafta, AI for War: Big Tech Empowering Israel’s Crimes and Occupation, Al-Shabaka: The Palestinian Policy Network, October 26, 2025.
[9] Jon R. Lindsay, US Military Leans Into AI for Attack on Iran, But the Tech Doesn’t Lessen the Need for Human Judgment In War, Conversation, March 11, 2026; Rich Duprey, Anthropic Deemed a ‘National Security Threat’ — Is Palantir Technologies At Risk? 24/7 Wallst.com.
[10] Dario Amodei, “Where things stand with the Department of War” Anthropic, March 5, 2026.
[11]Enrique Dans, “The Pentagon–Anthropic clash is a warning for every enterprise AI buyer,” Fast Company, March 12, 2026.
[12] Joshua Franklin, Akila Quinio and James Politi, “Scott Bessent called in US bank CEOs to discuss Anthropic model’s cyber risks,” Financial Times April 9, 2026.
[13] Samantha Subin, “Vance, Bessent questioned tech giants on AI security before Anthropic’s Mythos release,” CNBC, April 10, 2026.
[14] Jake Bleiberg and Margi Murphy, ‘White House Works to Give US Agencies Anthropic Mythos AI,” Bloomberg, April 16, 2026.
[15] Billy Perrigo, “Exlcusive: Anthropic Drops Flagship Safety Pledge,” Times February 24, 2026.
