OpenAI is expanding the deployment of its artificial intelligence tools across NATO and European government cybersecurity agencies, helping defenders identify software vulnerabilities, investigate potential threats and develop fixes as governments look to strengthen critical digital infrastructure.
In a Tuesday announcement, OpenAI said its partnership with NATO's Communications and Information Agency, or NCIA, provides the alliance's core institutions with access to advanced cybersecurity models designed to protect allied networks.
The initiative, powered by OpenAI's Daybreak program, is already seeing significant use within NATO, where the alliance has consumed as much as 300 billion raw input and output tokens per month through its work with the company.
NATO Tests AI Agents in Cybersecurity Exercises
The company also supported the 2026 Crossed Swords exercise, organized by the NATO Cooperative Cyber Defense Center of Excellence. The exercise introduced a new training track in which participants built, deployed and tested AI agents for cyber operations using Daybreak. The program emphasized human oversight, with cybersecurity professionals operating alongside the AI agents.
"AI-powered cyber defence models represent a generational leap in cyber security," said Dylan Browne, general manager at NCIA. He said the technology could accelerate vulnerability discovery, identify emerging threats and help NATO strengthen its cyber resilience.
Ukraine, EU Agencies Expand Access to OpenAI Models
OpenAI is also extending access beyond NATO, providing its cybersecurity models to the European Union Agency for Cybersecurity and national cyber agencies in France, Germany and Poland.
Ukraine's Ministry of Digital Transformation is using Daybreak to help protect civilian infrastructure, with the technology intended to accelerate the identification of software vulnerabilities and the development and testing of fixes amid persistent cyberattacks.
OpenAI highlighted several early results from the program. ENISA used its models to identify four vulnerabilities in software used across EU institutions, with fixes published for all four.
A November joint report from ENISA and CERT-EU also examined 27 software projects using frontier AI models, including OpenAI models accessed through Daybreak and its Codex security plug-in. Software maintainers confirmed 95% of 88 potential issues assessed during the review as genuine vulnerabilities.
CERT Polska Identifies 6 MikroTik Vulnerabilities
Separately, Poland's CERT Polska used OpenAI models to help identify six vulnerabilities in MikroTik RouterOS, coordinating disclosure with the software vendor. Researchers directed the process, tested findings on real systems and verified their impact, while AI helped accelerate the analysis.
MikroTik subsequently released fixes that CERT Polska confirmed prevented attacks observed by researchers, giving network operators a way to protect affected devices.
"The people defending our democracies should have the most advanced technology on their side," said Sasha Baker, OpenAI's head of national security policy.
OpenAI said it plans to continue working with governments and international cybersecurity partners to help defenders identify weaknesses, develop fixes faster and protect essential services.
