^{Getty Images}

OpenAI CEO Sam Altman moved to address internal concerns this week, telling employees that decisions about how artificial intelligence is deployed by the U.S. military ultimately rest with government officials — not with OpenAI.

During a company-wide meeting, Altman made it clear that while OpenAI can advise on technical capabilities and implement safety safeguards, it does not control operational decisions made by the Department of Defense. His remarks came days after the company announced an expanded partnership with the Pentagon, a move that surprised some employees and ignited debate across the AI industry.

Separation Between Technology and Military Strategy

Altman emphasized that OpenAI’s responsibility lies in building reliable models and implementing what he described as a strong “safety stack.” However, once the technology is deployed within government systems, strategic and operational decisions are determined by defense leadership.

The comments reflected a broader tension inside technology companies working with defense agencies. Employees at several major AI labs have expressed concern about how generative AI models could be used in military contexts, particularly in light of escalating geopolitical conflicts.

According to people familiar with the internal meeting, Altman said the Pentagon has shown respect for OpenAI’s technical expertise and has welcomed input on appropriate use cases. At the same time, the Department of Defense has been explicit that authority over military operations lies with senior government officials, including Defense Secretary Pete Hegseth.

Timing of the Defense Deal Raises Questions

The controversy intensified because OpenAI’s announcement of its expanded defense agreement came just hours before U.S. and Israeli strikes on Iran. The timing led to criticism that the company had moved too quickly to formalize the partnership.

Altman later acknowledged publicly that the rollout of the announcement “looked opportunistic and sloppy,” adding that the company should not have rushed to release the news. Still, he defended the substance of the agreement, stating that defense officials had demonstrated a serious commitment to safety and responsible integration of AI systems.

Under the new arrangement, OpenAI will be permitted to deploy its models across classified Defense Department networks. This marks a significant expansion from last year’s $200 million Pentagon contract, which allowed the military to use OpenAI’s systems in nonclassified environments.

Industry Rivalries and National Security Tensions

The deal also unfolded against the backdrop of escalating tensions involving rival AI lab Anthropic. The company was recently designated a national security supply-chain risk and subsequently blacklisted from certain federal uses. Federal agencies were instructed to halt deployment of its technology.

Anthropic had previously been the first AI lab to deploy models across the Defense Department’s classified network. Negotiations reportedly collapsed after disagreements over permissible use cases. Anthropic sought guarantees that its models would not be used in fully autonomous weapons systems or for mass domestic surveillance, while defense officials pushed for broader deployment authority across all lawful military applications.

The competitive dynamic adds another layer of complexity. xAI, founded by Elon Musk, has also entered agreements to make its models available for classified use cases. Musk and Altman, who co-founded OpenAI in 2015, are currently engaged in a high-profile legal dispute scheduled to go to trial next month.

During the internal meeting, Altman suggested that at least one competitor may be willing to accommodate government requests with fewer safety constraints. He reiterated his belief that OpenAI’s long-term advantage lies in building the most capable models while maintaining safeguards strong enough to preserve trust, even if those guardrails create friction with government partners.

Balancing Safety, Growth and Government Demand

The Defense Department’s interest in advanced AI systems has grown rapidly over the past two years, particularly for intelligence analysis, logistics optimization, cybersecurity defense and mission planning. AI integration into national security frameworks is accelerating, driven by competition with China and increasing reliance on data-driven decision-making.

For OpenAI, the Pentagon contract represents both opportunity and risk. Defense partnerships can generate significant revenue and institutional credibility, but they also invite scrutiny from employees, regulators and the public.

Altman’s message to staff underscored a central distinction: OpenAI can define the technical limits and safety mechanisms of its models, but it does not determine how sovereign governments conduct military operations. That boundary, he indicated, is fundamental to the company’s role as a technology provider rather than a policymaker.

As artificial intelligence becomes increasingly embedded in defense infrastructure worldwide, technology companies are facing difficult questions about responsibility, influence and accountability. OpenAI’s evolving relationship with the Pentagon illustrates how quickly AI innovation has moved from consumer tools to core national security systems — and how complex the governance challenges have become.