OpenAI signs agreement with the Department of War for classified-network use
Announced hours after Anthropic was barred from federal contracts, the deal let Pentagon use OpenAI's models 'for all lawful purposes'; Altman later called the timing 'rushed'.
- Government & policy
- Security & misuse
- Major
OpenAI announced an agreement with the US Department of War to deploy its models on the Pentagon’s classified networks, hours after President Trump ordered federal agencies to stop using rival Anthropic’s technology and Defense Secretary Pete Hegseth designated Anthropic a “supply-chain risk.” The contract permitted Pentagon use of OpenAI’s models “for all lawful purposes,” but included restrictions on mass surveillance of US citizens and on autonomous weapons systems operating without human decision-making — the same two guarantees Anthropic had refused to give up in the negotiations that led to its own exclusion.
OpenAI said it would enforce the restrictions both contractually, through language tied to existing laws and Defense Department policy, and technically, through prompt classifiers and fine-tuning intended to prevent the prohibited uses. The Department affirmed OpenAI’s services would not be used by Department of War intelligence agencies such as the NSA.
The timing drew immediate criticism, given the deal followed Anthropic’s blacklisting by less than a day and appeared to reward the company willing to accept terms its rival had rejected. Sam Altman later acknowledged in a public Q&A that the deal “was definitely rushed, and the optics don’t look good,” saying OpenAI had moved quickly because it was “genuinely trying to de-escalate” an increasingly public standoff between the military and Anthropic rather than to capitalise on a competitor’s exclusion. The episode intensified public debate over the two companies’ rival approaches to military contracting, with reporting at the time noting a spike in ChatGPT app uninstalls and a rise in Claude’s app-store ranking in the days that followed.