Sam Altman told OpenAI staff the company cannot dictate how the Pentagon uses its AI, saying employees "do not get to make operational decisions." The remarks followed reports that AI tools were used in U.S. operations related to Nicolás Maduro and strikes tied to Iran. Anthropic refused a Pentagon deal over surveillance and autonomous-weapons concerns and was labeled a "supply-chain risk" by Pentagon officials. A Pentagon pact with OpenAI sparked employee backlash; Altman admitted the announcement looked "opportunistic and sloppy."
Sam Altman: OpenAI Can’t Control Pentagon’s Use Of Its AI — Deal Prompts Employee Backlash
Sam Altman, OpenAI’s CEO, attends a session at the AI Impact summit at Bharat Mandapam in New Delhi, India, on 19 February 2026.Photograph: Rajat Gupta/EPA(Photograph: Rajat Gupta/EPA)
OpenAI CEO Sam Altman told employees that the company does not control how the Pentagon may apply its artificial-intelligence tools in military operations, according to reporting by Bloomberg and CNBC. The comments have intensified public debate about military uses of AI and triggered internal unrest among OpenAI staff.
What Altman Said
"You do not get to make operational decisions,"Altman said in the remarks cited by reporters. He added,
"So maybe you think the Iran strike was good and the Venezuela invasion was bad. You don’t get to weigh in on that."
Context And Controversy
The disclosure comes amid heightened scrutiny of how defense agencies are integrating AI. Reporters have said AI-enabled systems were used in some U.S. military operations, including an operation reported to involve Venezuelan leader Nicolás Maduro and in targeting-related decisions involving Iran. Those accounts have amplified ethical concerns from AI researchers and company employees about how models might be repurposed.
Anthropic, the maker of the Claude chatbot and a competitor to OpenAI, declined a Pentagon contract last week, citing fears its model could be used for domestic mass surveillance or fully autonomous weapons. In the wake of that refusal, Pentagon officials reportedly described Anthropic as a "supply-chain risk," a label that drew strong attention because it is uncommon for a U.S. technology firm.
OpenAI’s Deal And The Fallout
On the same day those reports emerged, the Pentagon announced a deal with OpenAI that appeared to substitute OpenAI services for programs that had considered Anthropic’s tools. The timing—and concerns that OpenAI accepted terms Anthropic rejected on ethical grounds—led to public criticism and frustration inside OpenAI.
Altman and company leaders moved to reassure employees and the public that they intend lawful use of their systems. Altman later conceded the announcement was rushed and acknowledged it made the company look "opportunistic and sloppy."
Rival CEO’s Response
Anthropic CEO Dario Amodei responded in an internal memo reported by The Information, calling Altman "mendacious" and accusing him of offering what Amodei described as "dictator-style praise to Trump." Amodei insisted Anthropic had maintained ethical red lines and criticized what he characterized as performative or superficial safety measures—"safety theater." He also suggested political and financial motives were at play, noting that OpenAI president Greg Brockman and his wife have given $25 million to a PAC supporting Donald Trump.
Why It Matters
The episode highlights unresolved questions about tech companies' influence over—or lack of control of—how their systems are used once sold to or contracted by government agencies. It also underscores tensions between ethical commitments from AI firms and national-security priorities that may demand relaxed safety constraints.
Key ongoing issues: whether AI vendors can set enforceable limits on military uses; how government pressure shapes company decisions; and how employees and executives reconcile safety commitments with defense contracts.
Help us improve.
