How Are AI Companies Contracting with the Department of War??
On February 27, 2026, OpenAI reached an agreement with the Department of War (DoW)—formerly known as the Department of Defense (DoD)—for “[deploying] advanced AI systems in classified environments,” which will likely be followed by similar agreements with other AI companies. OpenAI defended the agreement in a statement, writing that “the US military absolutely needs strong AI models to support their mission” as potential adversaries integrate AI technologies. The agreement was initiated after Anthropic cancelled a similar contract with the DoW, raising ethical concerns about the use of its AI systems in warfare.
After an internal re-evaluation of the agreement’s initial language and guardrails, OpenAI CEO Sam Altman posted on X that “[OpenAI] shouldn’t have rushed to get this out on Friday,” calling the initial terms “opportunistic and sloppy.” Altman assured Americans that they will continue to refine the agreement “with technical safeguards and other methods” in mind.
What are the New Conditions of the OpenAI-DoW Deal?
The agreement was updated on March 2nd, 2026, with a statement that OpenAI, as well as any other AI company working with DoW, and the DoW will operate under three major guidelines involving mass domestic surveillance, direct autonomous weapons systems, and high-stakes automated decisions without human control or operators. OpenAI stressed that their tools should not and will not be used to “conduct domestic surveillance” of American citizens. Additionally, OpenAI assured Americans that the company’s AI services will not be used by agencies such as the National Security Agency to collect or use “commercially-acquired personal or identifiable information.”
Before AI-assisted or AI-integrated tool services can be deployed, all services must comply with DoD Directive 3000.09. Issued in January 2023, the directive specifies that any use of AI “in autonomous and semiautonomous systems must undergo rigorous verification, validation, and testing” before deployment. Furthermore, AI-assisted or AI-integrated activities handling private information must comply with the Fourth Amendment, the National Security Act of 1947, the Foreign Intelligence and Surveillance Act of 1978, and Executive Order 12333. Together, these three judicial and legislative measures outline that AI systems can not be used “for unconstrained monitoring of U.S. persons’ private information.”
If the government violates any terms of the contract, then OpenAI has pledged to terminate its contract and services immediately, according to OpenAI’s public statement.
What do Tech Companies Think of AI-Powered Warfare?
On March 7, 2026, atop OpenAI robotics and hardware leader, Caitlin Kalinowski, resigned directly due to her concerns about the OpenAI-DoW deal. “AI has an important role in national security,” Kalinowski posted, “but surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got.”
Executives and leaders of other leading AI companies have also voiced concerns over the DoW’s use of AI systems for warfare, including the CEO of Anthropic. Less than one month before the OpenAI-DoW deal was announced, Anthropic pulled out of its $200 million contract with DoW in late February 2026 for ethical concerns. This contract would have allowed the Department to use Anthropic’s AI products in American surveillance and direct autonomous weapons without human oversight, unlike the OpenAI-DoW deal. Defense Secretary Pete Hegseth threatened to blacklist Anthropic’s services from government contracts and is “planning to keep using Anthropic’s tools” against Anthropic’s wishes, possibly invoking the Defense Production Act to force the company to produce its AI product for national security.
Last summer, DoW offered contracts worth up to $200 million to various AI frontier labs, including Google, OpenAI, xAI, and Anthropic. Because many of these contracts and deals are still being drafted and decided upon, OpenAI’s work with DoW is serving as a testing ground that other AI companies are watching closely to decide whether a DoW contract is a path worth taking.
Conclusion
Under the deal’s guardrails, AI systems utilized by the Department of War are not able to conduct mass surveillance, power autonomous weapons, or be deployed without limitations. OpenAI’s CEO supports the deal by arguing that technology offers a valuable opportunity to be ahead of critical threats and adversaries to national security, many of whom are already using AI. Pushback against the OpenAI-DoW deal reflects larger concerns about the use of autonomous systems in warfare, especially considering the speed at which the deal was developed.