News

OpenAI is giving NATO institutions access to its cybersecurity models

Agencies in Europe and Ukraine are already using the models through OpenAI's Daybreak program to find and fix security flaws.

Image: DepositPhotos

OpenAI announced that it is working with the NATO Communications and Information Agency (NCIA) to give the alliance’s core institutions access to its advanced cybersecurity models. Access runs through Daybreak, a program OpenAI launched in May, whose higher tiers are meant for vetted defenders doing authorized work such as reviewing code, investigating suspicious activity, and developing and testing fixes.

According to OpenAI, NATO has used up to 300 billion input and output tokens a month, though the company doesn’t say what period that figure covers. The announcement came during NATO Edge, the NCIA’s flagship technology event, held this year in Izmir, Turkey.

NCIA general manager Dylan Browne says that as adoption grows, the agency will build a better understanding of token consumption and the economics of AI models, which will help it cut costs, evaluate competing solutions and make informed procurement decisions. That suggests NATO is still testing the models rather than committing to OpenAI alone.

OpenAI also supported Crossed Swords 2026, the exercise run by the NATO Cooperative Cyber Defence Centre of Excellence (CCDCOE) in Tallinn, where a new training stream had participants build and test AI agents for cyber operations in Daybreak, with an emphasis on keeping humans in charge.

The company also points to examples from Europe. According to OpenAI, the EU cybersecurity agency ENISA used its models to find four vulnerabilities in software used across EU institutions, and fixes have been released for all four. A joint ENISA and CERT-EU report reviewed 27 software projects with several frontier models, OpenAI’s among them, and maintainers confirmed 95% of the 88 potential issues they assessed as real, a figure that covers all the models together.

Poland’s CERT Polska, according to OpenAI, used its models to help find six vulnerabilities in MikroTik RouterOS, the operating system on MikroTik’s network gear, and MikroTik has since released fixes. Ukraine’s Ministry of Digital Transformation uses Daybreak to defend civilian infrastructure, and France’s ANSSI and national agencies in Germany and Poland also have access.

This isn’t new territory. Anthropic introduced Project Glasswing in April to find flaws in critical software, while Google first gave Gemini 4 Argon and Gemini 4 Carbon only to cybersecurity experts.