Altman to OpenAI Staff: Hegseth Is in Driver's Seat

Sam Altman told OpenAI staff this week that when it comes to how the US military uses the company's AI, the Pentagon—not OpenAI—is in charge. In an all-hands meeting Tuesday, held four days after OpenAI revealed a new Defense Department deal, the CEO underscored the fact that the company won't be in charge of "operational decisions" regarding how the Pentagon uses its technology. "So maybe you think the Iran strike was good and the Venezuela invasion was bad. You don't get to weigh in on that," he said, according to a transcript seen by CNBC. OpenAI did change its agreement somewhat after backlash to the deal, stating domestic surveillance is not allowed, the Wall Street Journal reports.

According to a source, Altman said the Pentagon wants OpenAI's input on where its models fit and will let the company build the safety systems it thinks are necessary, but stressed that operational calls are solely the domain of Defense Secretary Pete Hegseth. The DOD deal, which expands a prior $200 million contract to include classified uses, has stirred criticism inside and outside OpenAI, especially after rival Anthropic was blacklisted by the Trump administration and cut off from federal use. Anthropic had sought limits on autonomous weapons and mass surveillance; talks with the Pentagon broke down.

Altman acknowledged that OpenAI's DOD rollout "looked opportunistic and sloppy" and shouldn't have been announced so quickly, but insisted the company maintains "a deep respect for safety." He suggested xAI—Elon Musk's firm, also working on classified networks—may offer the military far fewer safety constraints. He said he thinks xAI will tell the Pentagon, "We'll do whatever you want." Meanwhile, in a memo to employees, Anthropic CEO Dario Amodei accused Altman of giving President Trump "dictator-style praise," the Guardian reports.