AI Goes to War — OpenAI Policy Reversal and the Anthropic–Palantir–AWS Pact

2024 was the year generative AI shed its claim to be a "peaceful Silicon Valley tool" and was openly redefined as critical national-security infrastructure of the United States.

Two events—one in January, one in November—stood as bookends.

10 January — OpenAI Removes "Military and Warfare"

On 10 January 2024 (US time), The Intercept reported a quiet revision to OpenAI's usage policy.

The prior version had listed prohibited uses including:

Our services may not be used for: military and warfare …

The phrase "military and warfare" was no longer there.

OpenAI's statement said the policy had been "rewritten to be clearer and simpler", that weapons development and harm to people remained prohibited. But the blanket prohibition on "military and warfare" did not return.

In the days that followed, OpenAI began disclosing collaborations with the US Department of Defense and DARPA—on cybersecurity tooling, on a dialogue system for veteran suicide prevention. The stance that "non-lethal military use is fine" travelled from the policy document into practice.

7 November — Anthropic, Palantir, AWS, and the US Intelligence Community

On 7 November 2024, Anthropic announced on its blog a strategic partnership with Palantir Technologies and Amazon Web Services.

The substance was infrastructure to deliver Claude into the US intelligence and defence community—the CIA, DIA, NSA and the rest of what is collectively called the IC (Intelligence Community). Anthropic's Claude 3 and 3.5 would run on Palantir's classified-data platform, on AWS's classified-cloud regions (GovCloud / Top Secret cloud). The announcement named the use cases: "analysing complex data, pattern recognition, review of documents at speed, support for intelligence operations".

The agreement carried safety provisos—humans would remain the final decision-maker, lethal decisions would not be delegated to the machine—but the picture of the safety-flagged Anthropic becoming a contractor to national intelligence agencies registered widely.

A General Convergence

Across 2024 the same direction set in across the industry.

• Meta: in July, opened Llama models to military and national-security agencies.
• Google: revised its "AI Principles" so that the prohibition on weapons and surveillance work was softened.
• Microsoft: continued JWCC (the JEDI successor) DoD contract, extending OpenAI models into defence clouds.
• Scale AI: extended its Project Maven contract for battlefield image analysis with the US Army.

The Silicon Valley norm that had refused military work—epitomised by Google employees' 2018 protest against Project Maven—reversed almost completely in the post-ChatGPT competition.

What Changed

Technically, no leap. The same Claude, the same Llama, the same ChatGPT—different licences, different customers, different operating environments.

What changed was the industry's vocabulary. From "Responsible AI" to "Sovereign AI"; from "Safety" to "Security". This is not just wording. It reflects a shift in the norm governing AI development—from ethics toward national strategy.

Geoffrey Hinton, after receiving the Nobel Prize in October 2024, used his acceptance address to argue for a treaty banning autonomous weapons. As he acknowledged, individual warnings travel poorly through the gravitational fields of US–China competition and industry economics.

2024 was the year the generative-AI industry stopped talking about its intentions and showed its customers.