Economy

Add as a preferred source on Google

OpenAI says NATO uses AI tools to help defend allied networks

NATO used up to 300 billion input and output tokens a month in cybersecurity work with OpenAI, the company said. European cyber agencies also used its models to find software vulnerabilities.

Updated 2 min read
Data centre cooling plant.
Data centre cooling plant.

NATO used up to 300 billion tokens a month, units of data processed or generated by AI, in work to protect allied networks, OpenAI said in a publication on Nov. 17.

The AI company described work extending beyond NATO to European and Ukrainian cyber agencies. It described the work in “Strengthening NATO and allied cyber defenses with AI.” It gave examples of vulnerabilities found and fixed, without specifying the period covered by NATO’s usage figure.

OpenAI said the EU’s cybersecurity agency, ENISA, used its models to identify four vulnerabilities in software used across EU institutions. Fixes were published for all four, giving the agency examples of completed repairs alongside the broader testing of AI tools.

A November joint report by ENISA and CERT-EU covered checks of 27 software projects using advanced AI models, OpenAI said. The tools included OpenAI models available through Daybreak, its offering for authorized cybersecurity work, and the Codex security plug-in.

Software maintainers confirmed 95% of the 88 potential problems they assessed as genuine across the models tested, the company said, citing the joint report. That finding covered multiple models, and OpenAI did not give a separate confirmation rate for its own tools. The company said the results showed the value of combining AI tools with human expertise.

In Poland, CERT Polska used OpenAI models to help find six vulnerabilities in MikroTik RouterOS and coordinated disclosure with the vendor, the company said. Researchers directed the work, tested findings on real systems and checked their impact. MikroTik released fixes that CERT Polska confirmed blocked the attacks it had seen.

OpenAI said it works with NATO’s Communications and Information Agency to give NATO’s core institutions access to advanced cybersecurity models. The company described Daybreak’s uses as reviewing software, investigating suspicious activity, checking vulnerabilities and developing and testing repairs. It said the NATO work also helps build expertise among multinational teams protecting allied networks.

The company also supported the 2026 Crossed Swords exercise, organized by the NATO Cooperative Cyber Defence Centre of Excellence, OpenAI said. A new training component had participants build, deploy and evaluate AI agents for cyber operations using Daybreak. Participants worked alongside the agents, with training emphasizing human oversight.

Ukraine’s Ministry of Digital Transformation is using Daybreak to help defend civilian infrastructure, OpenAI said. The initiative aims to help teams find software weaknesses and develop and test fixes faster amid persistent cyberattacks. The company said it has also provided cyber models to national agencies in France, Germany and Poland, and would continue working with governments and international partners.

Related articles

View all