NATO's cyber defense arm now processes up to 300 billion tokens per month using OpenAI's Daybreak models, enabling allied agencies to scan software, validate fixes, and deploy AI agents for cyber operations in real time.
TL;DR
- NATO's cyber defense agency (NCIA) is running up to 300 billion tokens monthly through OpenAI's Daybreak models to strengthen allied cyber resilience.
- OpenAI's tools helped ENISA identify four vulnerabilities in EU software and CERT Polska find six flaws in MikroTik RouterOS, with fixes deployed by vendors.
- Daybreak is also powering NATO's Crossed Swords 2026 exercise, where participants built and tested AI agents for cyber operations under human oversight.
What happened
OpenAI's Daybreak cybersecurity models are now integrated into NATO's cyber defense workflow, with NCIA processing up to 300 billion tokens monthly to support vulnerability discovery and patch validation.
The models are used by NATO's Communications and Information Agency (NCIA) to protect Allied networks, marking one of the first large-scale deployments of frontier AI in government cyber defense.
During the 2026 Crossed Swords exercise, NATO's Cooperative Cyber Defence Centre of Excellence (CCDCOE) ran a training stream where participants built, deployed, and evaluated AI agents for cyber operations using Daybreak in a controlled environment.
Beyond NATO, OpenAI has extended Daybreak access to allied governments including Ukraine's Ministry of Digital Transformation, France's ANSSI, Germany's BSI, and Poland's CERT Polska, with early results showing measurable improvements in vulnerability detection and remediation.
Why it matters
For AI startups and developers, this deployment demonstrates how frontier models can be operationalized in high-stakes security environments, providing a blueprint for scaling AI in regulated sectors like defense and critical infrastructure.
The rapid identification and patching of vulnerabilities—such as the four flaws found in EU software and six in MikroTik RouterOS—shows how AI can reduce mean time to remediation, a critical advantage against fast-moving cyber threats.
The NATO exercise also highlights the importance of human oversight in AI-driven cyber operations, ensuring that automated agents augment rather than replace skilled defenders.
For investors, the partnership underscores the growing demand for AI tools tailored to government and enterprise security, with potential to drive adoption of specialized cybersecurity models across public and private sectors.
Key facts
- NATO's NCIA processes up to 300 billion tokens per month using OpenAI's Daybreak models for cyber defense operations.
- OpenAI supported NATO's 2026 Crossed Swords exercise, where participants built and tested AI agents for cyber operations using Daybreak under human oversight.
- The EU's ENISA used Daybreak to identify four vulnerabilities in software used across EU institutions, with fixes published for all four.
- CERT Polska used Daybreak to find six vulnerabilities in MikroTik RouterOS and coordinated their disclosure with the vendor, leading to patches that prevented observed attacks.
- Ukraine's Ministry of Digital Transformation is using Daybreak to strengthen cyber defense of civilian infrastructure amid persistent cyber attacks.
- A joint report from ENISA and CERT-EU reviewed 27 software projects using frontier models, confirming 95% of 88 potential issues assessed as genuine.
- OpenAI's Daybreak models are also accessible to national cyber agencies in France (ANSSI), Germany, and Poland.
Context
This deployment marks a significant step in the integration of frontier AI models into government cybersecurity, a trend accelerated by the increasing sophistication of state-sponsored cyber threats.
NATO's adoption of Daybreak follows broader efforts by allied nations to modernize cyber defense with AI, including the U.S. Cybersecurity and Infrastructure Security Agency's (CISA) AI-driven vulnerability scanning initiatives.
The use of AI in cyber defense is not limited to detection; it also enables faster patching and validation, reducing the window of exposure for critical systems and infrastructure.
