OpenAI’s Sam Altman addresses staff over Pentagon deal, says operational calls are up to the government

In a recent internal meeting held on March 3, OpenAI CEO Sam Altman addressed employees regarding the company's controversial agreement with the Pentagon. Altman emphasized that the operational aspects of using OpenAI's AI technology fall under the jurisdiction of the government, not the company itself. Citing a report from CNBC, Altman conveyed a clear message from the Pentagon, stating, "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 reiterated that the U.S. military intends to leverage OpenAI's expertise in determining how AI models should be utilized while permitting the organization to establish its own safety protocols as it deems appropriate. Importantly, Altman clarified that ultimate authority over operational decisions resides with Defense Secretary Pete Hegseth. He expressed optimism that OpenAI could develop superior models that would encourage government cooperation, even if some safety measures might be seen as inconvenient. Moreover, Altman referenced Elon Musk's xAI initiative, which aims to deploy models for classified applications, suggesting that xAI might be more accommodating to government requests. He stated, "But there will be at least one other actor, which I assume will be xAI, which effectively will say ‘We’ll do whatever you want.’" Since the announcement of the Pentagon contract, Altman has been actively defending the deal on social media platforms, particularly X (formerly Twitter). He admitted that the agreement was made in haste and acknowledged the Pentagon's commitment to safety and partnership for achieving optimal outcomes. This internal discussion comes against the backdrop of criticism surrounding OpenAI's contract announcement, which followed rival Anthropic being designated as a "Supply-Chain Risk" to national security. Notably, Anthropic is now re-engaging with the Pentagon regarding an AI contract after previously refusing to compromise on its AI safety measures related to domestic surveillance and autonomous weaponry.