DoD Picks OpenAI, Citing Stricter Anthropic 'Guardrails'

The OpenAI deal is part of a broader Pentagon strategy to accelerate AI adoption and become an "AI-first" warfighting force. This initiative, mandated by President Donald Trump and detailed in a January 2026 memo from Defense Secretary Pete Hegseth, aims to streamline data sharing, expand computing access, and partner with private-sector AI developers to maintain a strategic advantage. Central to the strategy are seven "Pace-Setting Projects," including "Swarm Forge" for testing AI-enabled combat concepts and "GenAI.mil," which provides department-wide access to generative AI models at classified levels. The Pentagon is also using special hiring authorities to recruit top AI talent and has directed all military branches to prioritize projects that fast-follow these initial efforts. OpenAI states its agreement includes more robust, multi-layered safeguards than any previous classified AI deployment. These protections are not just contractual but are built into the cloud-only deployment architecture, with cleared OpenAI personnel remaining in the loop and retaining control over the safety stack. The company has established three main "red lines": no use of its technology for mass domestic surveillance, to direct autonomous weapons systems where human control is required by law, or for high-stakes automated decisions that require human approval. The contract explicitly references current laws and policies, meaning that even if those change, the use of OpenAI's systems must align with the standards in place at the time of the agreement. This contract award follows a period of significant growth in the DoD's AI spending, which saw a 1500% increase in contract values from 2022 to 2023, reaching $4.3 billion and accounting for 95% of all federal AI funding. The Pentagon has been actively awarding large contracts to non-traditional firms, signaling a strategic shift in its procurement landscape. The situation echoes the 2018 controversy around Project Maven, where Google ultimately withdrew from the military contract after significant employee protests over the use of AI for analyzing drone footage. That incident sparked a broader debate in Silicon Valley about the dual-use nature of AI and the ethics of building military technology. Anthropic, for its part, has recently revised its own "Responsible Scaling Policy." The company removed a previous commitment to unconditionally halt AI development if its capabilities outpaced safety measures, citing a competitive and "anti-regulatory political climate" as a reason for the change. The Pentagon had previously awarded contracts of up to $200 million each to Anthropic, Google, OpenAI, and xAI to address national security challenges. However, the dispute with Anthropic escalated when the DoD insisted on the right to use AI for "all lawful purposes," a condition Anthropic found too broad, leading to the current standoff.