NATO has been using OpenAI’s cybersecurity models to help protect its networks, consuming as many as 300 billion input and output tokens a month, OpenAI said on November 17. The company said it works with the NATO Communications and Information Agency, or NCIA, to give the alliance’s core institutions access to the models.
The access runs through Daybreak, the program under which OpenAI offers its cyber models to vetted defenders for authorized work such as reviewing software, investigating suspicious activity, validating vulnerabilities, and developing and testing fixes. OpenAI split the program into two access tiers in August, when it released a model tuned to refuse fewer requests for offensive security work. Its post does not say which models NATO uses, when the work began or what NATO pays.
“AI-powered cyber defence models represent a generational leap in cyber security,” NCIA General Manager Dylan Browne said in a statement published by OpenAI. He said NATO would study the economics as its use grows. “As we scale adoption, we will build a deeper understanding of token consumption, usage patterns and the underlying economics of AI models,” he said. “This insight will help us optimise costs, evaluate competing solutions and make informed procurement decisions.”
Sasha Baker, OpenAI’s head of national security policy, said the company looked forward to “deepening this work with NATO and our allies.” The announcement came as NATO Edge 26, the NCIA’s conference with industry, opened in Izmir, Turkey.
OpenAI also supported Crossed Swords 2026, the annual exercise run by the NATO Cooperative Cyber Defence Centre of Excellence in Tallinn. The center said the exercise, held from November 2 to 6, brought together more than 250 participants from over 40 NATO and non-NATO countries. A new training stream had participants design, develop, deploy and evaluate AI agents for cyber operations while working alongside them as operators. The technical environment used on-premises AI and, for the first time, Daybreak.
“This year, that gets harder, because AI is now in the loop on both sides,” exercise director Jozef Kostelansky said of synchronizing cyber effects with kinetic and special operations actions. He said participants were being trained to use AI “while maintaining the critical human oversight required for complex, multi-domain mission success.”
Beyond NATO, OpenAI said it has given access to its cyber models to the European Union’s cybersecurity agency, ENISA, and to national cyber agencies in France, Germany and Poland. It said Ukraine’s Ministry of Digital Transformation is using Daybreak to defend civilian infrastructure. ENISA used the models to identify four vulnerabilities in software used across EU institutions, with fixes published for all four, OpenAI said, and Poland’s CERT Polska used them to help find six vulnerabilities in MikroTik’s RouterOS. A joint report from ENISA and CERT-EU, as OpenAI describes it, reviewed 27 software projects using frontier models and found that 95 percent of the 88 potential issues assessed by maintainers were genuine.
The announcement came a day after California’s attorney general sued OpenAI, alleging it tested AI models it knew to be dangerous without the safeguards and monitoring needed to protect the public. OpenAI has said it committed $1 billion in subsidized Daybreak access over six months for defenders of essential services and digital infrastructure.
Sources: OpenAI, NATO CCDCOE, UK Defence Journal
–
By the Control Plane Editorial Team