Guess who the US military just recruited? Private AI

On 12 January, Pete Hegseth, an ex-TV “personality” with big hair who is now the US secretary for war (nee defence), bounded on to a podium in Elon Musk's SpaceX headquarters in Texas. He was there to announce his plans for reforming the American war machine’s bureaucratic engine, the Pentagon. In a long and surprisingly compelling speech, he made it clear that he’s embarked on a radical effort to reshape the bureaucracy of the war department, and to break up its cosy relationships with what Dwight Eisenhower called the “military-industrial complex” – the handful of bloated defence contractors that have assiduously milked the US government for decades while never delivering anything that was on time and within budget.

Predictably, one of the tools that Hegseth had chosen for his demolition job was AI, and to that end, three companies – Anthropic, Google and OpenAI – had already been given $200m contracts by the Pentagon to develop AI “agents” across different military areas. Given the venue and his host for the day, it came as no surprise to those present when Hegseth announced that Musk’s AI model, Grok, was also going to be deployed on this radical mission.

This did come as a surprise, though, to those outside the SpaceX hangar. Did it mean, mused the mainstream media commentariat, that this AI tool, which was mired in outrage and controversy for enabling people to create sexualised images of children, would be empowered to roam freely through all the archives – classified as well as unclassified – of the US war department?

Answer: yes, since it was part of “a transformative artificial intelligence acceleration strategy that will extend our lead in military AI deployment and establish the United States as the world’s undisputed AI-enabled fighting force”, according to a war department press release. “Mandated by President Trump, this acceleration strategy will unleash experimentation, eliminate legacy bureaucratic blockers, and integrate the bleeding edge of frontier AI capabilities across every mission area to usher in an unprecedented era of American military AI dominance.”

No holds will be barred in this, er, crusade. “Gone are the days of equitable AI,” declared the secretary for war, “and other DEI [diversity, equality and inclusion] social justice infusions that constrain and confuse our employment of this technology. Effective immediately, responsible AI at the war department means objectively truthful AI capabilities, employed securely and within the laws governing the activities of the department. We will not employ AI models that won’t allow you to fight wars.”

Stirring stuff, eh? But it makes you wonder what Anthropic, maker of the Claude AI model, is doing getting mixed up in this crusade. After all, it’s always prided itself as an outfit that takes ethics seriously. It’s a private firm, but structured as a public benefit corporation, explicitly prioritising safety alongside profit. It has a conservative deployment philosophy, releasing fewer models, iterating more cautiously and deferring release of new capabilities until it believes safety mitigations are ready.

So is Anthropic’s involvement in the Hegseth strategy really its first big step on the slippery slope that leads most tech companies into the military-industrial complex? The standard case study here is Google, which, aeons ago, was a very high-minded outfit. “Don’t be evil” was its motto when it launched as a public company in 2004.

In 2018, thousands of its employees signed a petition calling on the company to end its work for the Pentagon on Project Maven, a pilot programme meant to speed up the military’s use of AI. “We believe that Google should not be in the business of war,” the letter stated. “Therefore we ask that Project Maven be cancelled, and that Google draft, publicise and enforce a clear policy stating that neither Google nor its contractors will ever build warfare technology.” Google responded by announcing that it would not seek to renew the contract with the Pentagon after it had expired.

Now spool forward to 9 December 2025. The war department announces “the launch of Google Cloud’s Gemini for government as the first of several frontier AI capabilities to be housed on GenAI.mil, the department’s new bespoke AI platform. This initiative cultivates an ‘AI-first’ workforce, leveraging generative AI capabilities to create a more efficient and battle-ready enterprise.”

The road to hell is indeed paved with good intentions.

Working relationship

On Writers and Their Day Jobs is a lovely essay by Ed Simon on the jobs writers have to do to enable them to write.

Royal visit

An insightful Substack post by Michael Ignatieff is When the Sun King Goes to Davos.

The full Danish

Your Primer on Greenland is knowledgable common sense by James Fallows.