OpenAI puts frontier models in NATO cyber defenders' hands: core institutions have used up to 300 billion raw tokens a month
Synopsis
OpenAI is working with NATO's Communications and Information Agency (NCIA) to give NATO's core institutions access to advanced cybersecurity models for reviewing software, investigating suspicious activity, validating vulnerabilities, and developing and testing fixes. NATO has used up to 300 billion raw input and output tokens per month in this work. The same models also serve European defenders: the EU's ENISA used them to identify four vulnerabilities in software used across EU institutions, with fixes published for all four, and CERT Polska used OpenAI models to help identify six vulnerabilities in MikroTik RouterOS, whose fixes CERT Polska confirmed prevented the attacks it had observed.
Interpretation
Frontier AI models are now placed directly in the hands of trusted national and alliance cyber defenders for authorized security work, from reviewing software and investigating suspicious activity to validating vulnerabilities and developing and testing fixes. Daybreak is the delivery entry point for this authorized security work, handing model access to defenders rather than having defenders build models themselves. NATO has used up to 300 billion raw input and output tokens per month as part of this work, a volume that moves the collaboration from intent to operating practice.
The collaboration takes the form of OpenAI providing NATO's core institutions with access to advanced cybersecurity models to help protect Allied networks. The counterpart is NATO's Communications and Information Agency (NCIA), and the covered objects are NATO's core institutions and Allied networks. The users are NATO's core institutions reached through NATO's Communications and Information Agency (NCIA), and the protected object is Allied networks.
The same models produced checkable vulnerability findings in European defensive work, closed by human review and vendor fixes. ENISA used the models to identify four vulnerabilities in software used across EU institutions, and fixes were published for all four. A November joint report from ENISA and CERT-EU describes a review of 27 software projects using frontier models, including OpenAI models accessed through Daybreak, and tools including the Codex security plug-in; across the models tested, 95% of the 88 potential issues assessed by software maintainers were confirmed as genuine.
Perspective
The demonstrated scope is authorized use of advanced cybersecurity models by NATO's core institutions, the 2026 Crossed Swords exercise in which participants built, deployed and evaluated AI agents for cyber operations, and concrete vulnerability identification and remediation work by defenders including ENISA, CERT Polska and Ukraine's Ministry of Digital Transformation. The intended reader is a researcher or engineer following how government and alliance-level cyber defense brings in frontier models.
The material is presented through the collaborators' and agencies' own accounts; the vulnerability counts and the 95% confirmation rate are tied respectively to the software ENISA reviewed, CERT Polska's work on MikroTik RouterOS, and the 88 potential issues assessed by maintainers, with no unified cross-agency evaluation. The exercise account states only that an agent training stream was added with emphasis on human oversight, and reports no exercise results.
