Trending Articles
Trending Topics
Trending
Most of us ignore the terms of service when signing up for a new digital product, but we aren’t using AI for wars.
A dispute between Anthropic and the Trump administration has sparked a battle for the soul of AI – assuming it has one – and shone a light on how the technology is already being used in conflicts.
So far, Anthropic has been ditched by the US Government while still being used for targeting in “Operation Epic Fury”, the war with Iran. This left a wide open space for OpenAI to step into. Although, as we shall see, its CEO Sam Altman claims OpenAI has its own line in the sand when it comes to using AI models in war.
This being the US in 2026, it’s also a war of words. Trump has called Anthropic everything from “woke” to “dogs”, while Anthropic Co-Founder and CEO Dario Amoedi compared the US President to a dictator. Something he has since apologised for.
Here, we ignore the mudslinging and attempt to search out the facts in the very unique war between Anthropic and Trump.
It all started in July 2025 when the Pentagon doled out contracts worth up to $200 million to AI developers. This included Anthropic, plus Google, OpenAI and xAI. However, Anthropic quickly became the only LLM-maker on classified networks, according to Reuters.
The company agreed to the use of its models with two conditions: it couldn’t be used to surveil American citizens or be used in a lethal weapons system without human oversight.
Then, in January this year, the American military captured Venezuelan President Nicolas Maduro. Reports later suggested Anthropic’s Claude was used to analyse intelligence.
Perhaps feeling keen after that success, in mid-February Defense Secretary Hegseth dropped an AI memorandum that said any Department of War (DoW) contracts had to reference “any lawful use” – in short, Anthropic’s conditions were no longer allowed.
Things kicked off when Anthropic refused to sign.
Hegseth gave Amodei a deadline of 5pm on Friday 27 February to agree to the new terms or face blowback. Specifically, losing its share of the $200 million contract and being banned from government use.
Amodei made it clear via a detailed statement the day before the deadline that his company wouldn’t comply, and acknowledged that would come with punishment. “In a narrow set of cases, we believe AI can undermine, rather than defend, democratic values,” the statement said. “Some uses are also simply outside the bounds of what today’s technology can safely and reliably do.”
That sparked a stern response from Emil Michael, the Undersecretary of Defense for Research and Engineering, who called Amodei “a liar” on social media, adding that he “has a God-complex” and wants to “personally control the US military”.
By this point Amodei and Michael had already met to discuss the contract, although the New York Times noted they very much dislike each other. No kidding.
In response to Google’s Chief Scientist Jeff Dean backing its rival on the surveillance issue, a Pentagon spokesperson echoed Emil Michael’s post by writing this on X on the Thursday before the deadline: “We will not let ANY company dictate the terms regarding how we make operational decisions. They have until 5:01 PM ET on Friday to decide. Otherwise, we will terminate our partnership with Anthropic and deem them a supply chain risk for DoW.”
And that’s what happened.
In response, Trump said on social media, in all caps: “The United States of America will never allow a radical left, woke company to dictate how our great military fights and wins wars!” xAI Founder Elon Musk weighed in with “Anthropic hates Western Civilization”. Trump later added he “fired” the company “like dogs” – an insult frequently used by the president, who is clearly more of a cat person.
The US Government has given itself six months to remove Anthropic from its systems. Days later it said it would designate the company a supply chain risk, which means no part of the government would work with companies that work with Anthropic. That, of course, means Anthropic risks losing further business.
Anthropic said it would challenge such a risk designation in court.
OpenAI CEO Sam Altman openly supported his rival – as did others across Big Tech – but within hours of the deadline passing his company had signed up to fill the gap filled by Anthropic’s looming removal from the US government. However, OpenAI said it had three rules in its contract: banning domestic surveillance, autonomous weapons systems or other important automated decisions.
“We think our agreement has more guardrails than any previous agreement for classified AI deployments, including Anthropic’s,” OpenAI said the Saturday after the deadline, though Altman later admitted the partnership looked “opportunistic and sloppy”.
Indeed, many were unconvinced. An open letter signed by several hundred OpenAI and Google employees complained about the deal, while a user-led boycott called for people to “quit ChatGPT” and ditch their subscriptions. So far, 2.5 million people have signed up on QuitGPT.org, which was originally set up to complain about OpenAI execs’ financial contributions to Trump. After which the Claude app becoming the most downloaded app on the App Store.
In a leaked memo, Amodei said Anthropic was disliked by the US Government because its execs don’t donate to Trump’s party and haven’t given “dictator-style praise” to Trump – noting OpenAI’s Sam Altman has. He has since apologised for his tone.
What’s next? First, Anthropic will sue over the supply chain designation. It will then hope to win the branding war to become the ethical AI company of choice, while its AI rivals will attempt to keep working with the US without letting their models be used to kill or stalk US citizens. We’ll also see if a mass boycott is enough to hurt OpenAI’s finances.
The day after the Anthropic deadline, something else happened: the US dropped bombs on Iran.
Reports suggest the US is using Anthropic’s Claude despite the disagreement and ensuing ban on the models. Sources told the Wall Street Journal that US Central Command in the Middle East was using Claude, and that generally it was used for assessing intelligence, simulating battle scenarios and identifying targets.
None of those use cases necessarily come up against the terms that sparked the divorce between Anthropic and the US Department of War. However, they highlight how reliant the military already is on AI, even when it’s angry with the AI model’s makers.
So far, AI’s use in war is less about killer robots and more about sifting through data to let humans make decisions – but it’s early days in the shift to AI wars. “AI is changing the nature of modern warfare in the 21st century. It is difficult to overstate the impact that it has and will have,” Craig Jones, a Senior Lecturer in Political Geography from Newcastle University, told Sky News. “It is a potentially terrifying scenario.”
