Article
OpenAI discloses that NATO uses its cybersecurity models, with monthly usage reaching up to 300000000000 tokens
BlockBeats news, November 17, according to The Defiant, OpenAI disclosed that it is providing access to advanced cybersecurity models to core NATO institutions through the NATO Communications and Information Agency (NCIA) to help protect allied networks. OpenAI stated that NATO's monthly raw input and output tokens reach as high as 300000000000. The company also supported the CCDCOE-organized Crossed Swords exercise in 2026, where participants used the defensive tool Daybreak to build and evaluate AI agents for cyber operations, emphasizing the importance of maintaining human oversight. NCIA General Manager Dylan Browne said these models can accelerate vulnerability discovery, detect emerging threats at machine speed, and help protect critical digital infrastructure.
In addition to NATO, OpenAI stated that it has provided access to its network models to the European Union Agency for Cybersecurity (ENISA), as well as to France's ANSSI, Germany's national cyber agency, and Poland's national cyber agency. Ukraine's Ministry of Digital Transformation is also using Daybreak to strengthen cyber defenses for civilian infrastructure. ENISA used relevant models to discover 4 vulnerabilities in software used by EU institutions, all of which have been patched; in a joint report with CERT-EU in November, they stated that among 27 software projects reviewed, 95% of the 88 potential issues assessed by maintainers were confirmed as genuine problems. Poland's CERT Polska leveraged OpenAI models to discover 6 vulnerabilities in MikroTik RouterOS and coordinated their disclosure, after which MikroTik released patches.