OpenAI publishes National Security Principles as it expands government cyber partnerships

OpenAI published a formal set of "National Security Principles" governing how it works with governments and national-security customers, disclosed alongside an expanding roster of cyber-defense and biosecurity partnerships with the U.S. and allied governments.

What's new

• "Today we are publishing OpenAI's National Security Principles to offer transparency into how we approach government partnerships and national security uses of our technology," OpenAI said, describing the principles as applying to its "current and future national security and law enforcement partnerships."
• The principles were developed with outside national-security expert David Kris, who OpenAI says helped facilitate the process and provided independent judgment, alongside internal listening sessions spanning research, safety, policy, and government-partnerships teams.
• OpenAI disclosed that, in the past month, its Daybreak cyber-defense program established "Trusted Access for Cyber" partnerships with Australia, Canada, Japan, South Korea, France, Germany, Poland, the Netherlands, and EU institutions including ENISA, plus a growing cyber partnership with the UK government.
• The company says it is taking a similar approach in biosecurity, citing last month's expanded trusted access to its GPT-Rosalind model for select U.S. government and allied partners supporting public-health and biodefense missions.

Context

The announcement lands as governments increasingly deploy frontier AI for sensitive national-security work, and as scrutiny grows over how the most capable models should be governed in defense and law-enforcement contexts. It follows the same week's news that the Trump administration lifted export restrictions tied to OpenAI's technology for more than 100 U.S. companies and agencies, and positions OpenAI alongside Anthropic and Google DeepMind in publishing explicit governance frameworks for state and defense customers rather than handling those relationships case-by-case.

Why it matters

A public, written commitment to "democratic accountability" and "meaningful human judgment" gives OpenAI a defensible framework to point to as it pursues larger government contracts, while getting ahead of safety and civil-liberties criticism before deals are announced rather than after. More concretely, the parallel expansion of Trusted Access cyber partnerships across nine countries plus EU institutions shows OpenAI moving from research-paper-level safety commitments to operational deployment of its models inside allied governments' cyber-defense infrastructure at meaningful scale.