TL;DR
OpenAI CEO Sam Altman told employees the company does not control how the Pentagon uses its AI systems, stating “you do not get to make operational decisions.” The admission came as the Pentagon demanded AI companies remove safety guardrails, and after Altman himself described the military deal as “rushed” and “opportunistic and sloppy.”
What Altman Said
In communications with OpenAI staff, Altman was direct about the limits of the company’s influence once its technology enters military hands. “You do not get to make operational decisions,” he told employees, drawing a clear line between providing AI tools and dictating how those tools are used in the field.
Altman also acknowledged shortcomings in how the deal came together. He described the Pentagon agreement as “rushed” and “opportunistic and sloppy” — a rare public admission of process failures from the CEO of the world’s most prominent AI company.
How the Deal Came About
The Pentagon announced its OpenAI agreement on the same day it effectively blacklisted rival Anthropic. Defence Secretary Pete Hegseth declared Anthropic a “supply-chain risk” after the company refused to accept contract terms that would have permitted autonomous weapons deployment and mass domestic surveillance.
Anthropic CEO Dario Amodei responded with a blistering memo that called Altman “mendacious” and accused the Pentagon of pressuring AI companies to strip away safety restrictions. The Pentagon had been demanding that AI suppliers remove guardrails from their models as a condition of defence contracts.
The timing — announcing the OpenAI deal simultaneously with Anthropic’s exclusion — suggested the arrangement was driven more by political dynamics than careful procurement planning.
Internal and External Pressure
Altman’s candid remarks to staff reflect growing tension inside OpenAI. Employees who joined the company based on its original safety-focused mission are reportedly uncomfortable with the speed and terms of the military partnership.
Outside the company, the backlash has been measurable. ChatGPT uninstall rates surged following the Pentagon announcement, while Anthropic’s Claude app climbed app store rankings as users voted with their downloads.
Looking Forward
Altman’s admission that OpenAI cannot control military deployment of its AI raises a fundamental question for the industry: once a model is sold, who bears responsibility for how it is used? For UK policymakers and defence procurement officials watching from across the Atlantic, the answer to that question will shape how Britain structures its own AI partnerships with both domestic and allied military forces.