NATO expands AI support for cyber defence through OpenAI Daybreak

NATO cyber teams gain access to Daybreak
OpenAI is working with NATO’s Communications and Information Agency (NCIA) to give NATO’s core institutions access to advanced cybersecurity models through Daybreak. The effort is intended for authorised security work, including software review, investigation of suspicious activity, vulnerability validation, and the development and testing of fixes.
NATO has used up to 300 billion raw input and output tokens per month in this work. OpenAI said the deployment supports teams responsible for protecting Allied networks and contributes lessons for defenders across Europe and the wider Alliance.
Dr Dylan Browne, NCIA General Manager, described AI-powered cyber-defence models as a generational leap for security work. He said that, as adoption scales, NATO will develop a deeper understanding of token consumption, usage patterns and the economics of AI models to optimise costs, assess competing options and inform procurement decisions.
Training keeps people in the operating loop
OpenAI also supported the 2026 Crossed Swords exercise, organised by the NATO Cooperative Cyber Defence Centre of Excellence. The exercise introduced a training stream in which participants built, deployed and evaluated AI agents for cyber operations in a technical environment using Daybreak.
The programme positioned participants alongside AI agents as operators and emphasised human oversight. OpenAI said it remains committed to NATO’s principles for responsible AI adoption and to providing safe and secure solutions.
European agencies report vulnerability findings
Beyond NATO, OpenAI has provided access to its cyber models to European defenders, including ENISA and national cyber agencies in France, Germany and Poland. Ukraine’s Ministry of Digital Transformation is using Daybreak to help teams identify software vulnerabilities and develop and test fixes for civilian infrastructure.
ENISA used the models to identify four vulnerabilities in software deployed across EU institutions, and fixes were published for all four. A joint November report by ENISA and CERT-EU reviewed 27 software projects using frontier models, including OpenAI models accessed through Daybreak and the Codex security plug-in.
Software maintainers confirmed 95% of the 88 potential issues assessed in that review as genuine. Separately, CERT Polska used OpenAI models to help identify six vulnerabilities in MikroTik RouterOS, coordinated disclosure with the vendor, and verified that released fixes prevented the attacks it had observed.
What organisations can take from the programme
The reported work places AI within a process led by security researchers, who test findings on real systems and verify their impact. For organisations assessing AI for defensive security, the practical implication is to use models to accelerate analysis while retaining expert validation, coordinated remediation and human oversight.

