OpenAI Publishes National Security Principles for Government AI Use

Governments are beginning to use frontier AI systems for increasingly important national security work, including cyber defense and biological security. This creates a new imperative for AI labs, governments, and civil society to collaborate on how these tools should be deployed in sensitive settings. OpenAI believes democratic societies should leverage AI to protect people and critical infrastructure, but only in ways that reinforce democratic accountability, meaningful human judgment, and the rule of law. The company published its National Security Principles to offer transparency into its approach.

The National Security Principles were developed through a cross-company effort, facilitated by national security expert David Kris, and include listening sessions across teams from research to policy. They apply to current and future partnerships, including existing work with the Department of War, and include contractual restrictions against mass domestic surveillance, autonomous weapons systems, and high-stakes automated decisions. OpenAI is expanding partnerships with the U.S. and allied governments in critical defensive areas. In the past month, through the Daybreak cyber defense program, it established Trusted Access for Cyber partnerships with Australia, Canada, Japan, South Korea, France, Germany, Poland, the Netherlands, and EU institutions. It also has a growing partnership with the UK government on cyber testing and evaluation. In biosecurity, OpenAI announced expanded trusted access to its GPT‑Rosalind model for select U.S. government and allied partners supporting public health and biodefense missions.

For serious builders, the key insight is that OpenAI is treating national security partnerships as a domain where the company’s role is to inform the democratic process, not to make decisions alone. The company supports legislative efforts to establish safeguards around the highest-risk military uses of AI, including domestic surveillance and autonomous weapons. This is a practical example of how an AI lab can codify its ethical boundaries through explicit principles and contractual restrictions, while acknowledging that the most consequential questions should be answered through democratic debate. It also shows the operational reality of expanding government partnerships in sensitive areas like cyber and biosecurity.