OpenAI is scaling its defense contracting operations by granting NATO and key European governments operational access to its Daybreak cybersecurity platform. Through the platform, defense teams deploy the underlying models to audit institutional codebases, map vulnerabilities, and patch software across government networks.
Automated Vulnerability Detection During the NATO exercise Crossed Swords 2026, military units deployed autonomous AI agents for cyber operations under human supervision for the first time. The deployment follows operational benchmarks from public agencies using the models for automated code audits. The EU cybersecurity agency ENISA deployed the models to isolate four vulnerabilities in institutional software. Furthermore, a joint assessment with CERT-EU analyzed 27 software projects and verified that 95 percent of the 88 reported anomalies represented actionable security bugs.
95 percent of the 88 reported issues turned out to be real bugs.
National security organizations are moving toward scaled deployment of these auditing frameworks. The NATO Communications and Information Agency (NCIA) has embedded the underlying models into daily operations, with NCIA chief Dylan Browne characterizing the shift as a structural transition in cyber defense capabilities. Ukraine is utilizing Daybreak to secure civilian infrastructure, while France, Germany, and Poland have secured administrative access. In Poland, CERT Polska leveraged the tooling to isolate six active vulnerabilities in exploited MikroTik routing hardware.
Strategic Shift in Software Infrastructure As government contracts expand, the financial terms and integration costs for deploying proprietary LLMs into classified public IT budgets remain undisclosed by procurement offices.
The catch
National security infrastructure now carries direct exposure to a single commercial model as defense alliances hand governmental network audits over to Daybreak. In Poland, CERT Polska pinpointed six active vulnerabilities inside compromised MikroTik network hardware using this setup.
Our stories: OpenAI Daybreak Cuts Check Point Threat Research to Minutes (October 28).