OpenAI chief Sam Altman told staff on Tuesday that the firm does not dictate how the Pentagon employs its artificial‑intelligence tools in combat settings. His statement about OpenAI’s lack of influence arrives as lawmakers and ethicists intensify scrutiny of military AI use and as engineers voice worries about the fate of their creations.
“You don’t get to make operational choices,” Altman said to employees, according to Bloomberg and CNBC.
“So whether you think the Iran strike was justified or the Venezuela operation was wrong, you’re not asked to weigh in,” he added.
The sector has been caught in fierce debate and strained talks in recent weeks after the Defense Department pressed AI firms to lift safety constraints on their models to broaden possible military uses. Sources say AI‑driven systems have already been deployed in the U.S. effort to detain Venezuelan leader Nicolás Maduro and in targeting decisions during the conflict with Iran.
Anthropic, a competitor that builds the Claude chatbot, turned down a Pentagon contract last week, citing fears that its model could be repurposed for domestic mass surveillance or fully autonomous weapons. Defense Secretary Pete Hegseth labeled the company a “supply‑chain risk,” a designation never before applied to a U.S. firm and one that could inflict serious financial damage if formalized.
On the same day Hegseth announced punitive steps against Anthropic, the Pentagon revealed a separate agreement with OpenAI that appeared intended to substitute Claude in military projects. The timing of that pact and the perception that OpenAI had crossed ethical lines rejected by Anthropic sparked criticism both publicly and within OpenAI’s workforce.
Altman and OpenAI have since stressed that their technology will be used lawfully and have sought to manage the fallout, with Altman conceding that the agreement was hurried and made the company appear “opportunistic and sloppy.”
Anthropic chief executive Dario Amodei denounced Altman as “mendacious” in a memo to staff on Wednesday and accused him of offering “dictator‑style praise to Trump,” according to The Information. “We have kept our red lines intact rather than colluding to stage ‘safety theater’ for the benefit of employees—something everyone at the Pentagon, Palantir, our political advisers, and others assumed we were trying to solve,” Amodei wrote.
He also targeted the Pentagon and the Trump administration. “The real reason the Pentagon and the Trump team dislike us is that we haven’t donated to Trump, whereas OpenAI and Greg have given heavily,” he added, referencing Greg Brockman, OpenAI’s president, who contributed $25 million to a pro‑Trump political action committee with his wife.
Read next
Starmer issues ultimatum to tech companies to prevent explicit content on children's devices
Prime Minister Keir Starmer announced on Monday that Apple and Google have until September to implement software that blocks explicit imagery on children's mobile devices, or face new legislation.
The prime minister stated that tech firms must employ nudity-detection algorithms or similar technical measures on tablets and smartphones.
Study finds AI self‑replicating in the wild, a first.
Recent research shows that some AI systems can now duplicate themselves onto other computers without human help, a capability that sounds like a scene from a sci‑fi film or an excited corporate blog post. In a worst‑case picture, a rogue super‑intelligent AI could avoid being shut down
European AI translation sector warned that partnering with US firms could harm its reputation
AI firms in Europe could lose their leading position in machine translation after one of the continent’s top startups decided to work with Amazon’s cloud division, prompting concern across the industry.
Although European businesses have generally trailed the United States and China in adopting artificial intelligence, a handful