NATO's NCIA Uses Up to 300B OpenAI Tokens a Month via Daybreak
TL;DR
- ENISA found 4 real vulnerabilities in EU software and Poland's CERT flagged 6 MikroTik RouterOS flaws, with 95% of 88 assessed issues confirmed genuine.
- Ukraine's digital ministry is running Daybreak against live Russian attacks, making this simultaneously a wartime and peacetime alliance deployment.
- NCIA will study token consumption, usage patterns, and competing solution economics before committing to procurement, leaving the vendor door open.
NATO's Communications and Information Agency has used up to 300 billion raw input and output OpenAI tokens a month to defend Allied networks, with access routed through Daybreak, OpenAI's program for vetted defenders doing authorized code review, suspicious-activity work and fix testing, Help Net Security reports. The same tier now sits behind cyber-defense work at Ukraine's Ministry of Digital Transformation, France's ANSSI, national agencies in Germany and Poland, ENISA and CERT-EU.
The concrete yield is modest but documented. ENISA used OpenAI models to find four vulnerabilities in software run across EU institutions, and fixes are out for all four. Together with CERT-EU, it reviewed 27 software projects using several frontier models, OpenAI's among them, along with tools such as the Codex security plug-in; maintainers assessed 88 potential issues and confirmed 95% of them as real. CERT Polska used OpenAI models to help find six vulnerabilities in MikroTik RouterOS; its researchers directed the work and checked findings on real systems. MikroTik released fixes, and CERT Polska confirmed they stop the attacks it had seen.
"As we scale adoption, we will build a deeper understanding of token consumption, usage patterns and the underlying economics of AI models. This insight will help us optimise costs, evaluate competing solutions and make informed procurement decisions," said NCIA general manager Dr Dylan Browne.
That procurement line matters because the volume is already at hyperscaler-grade scale while the economics are not. The disclosure also lands inside a thick run of related coverage: it joins 406 OpenAI stories we have tracked in the past 90 days, including a California AG suit filed Monday over the Hugging Face breach.
At NATO's 2026 Crossed Swords exercise, run by the CCDCOE, participants built and tested AI agents for cyber operations inside Daybreak, with an emphasis, the article says, on keeping humans in charge.
What others are reporting
-
OpenAI Read →
First-party post naming France, Germany, Poland, and Ukraine as participants; sets program constraints: authorized systems only, no external access sharing.
NATO will study token consumption, usage patterns and underlying economics of AI models before making procurement decisions.
-
UK Defence Journal Read →
UK defence-specialist framing; covers Crossed Swords 2026 in Tallinn and Ukraine's minister citing AI making Russian attacks more massive and cheaper.
AI-powered cyber defence models represent a generational leap in cyber security.
-
The Decoder Read →
Leads with granular operational evidence: ENISA's 4 bug finds, Poland CERT's 6 MikroTik vulns, and the 95% confirmation rate across 88 assessed issues.
-
The Defense News Read →
Emphasizes human-in-the-loop verification protocols and frames the 300B token figure as an operational scale milestone rather than a partnership announcement.
These capabilities help security teams find weaknesses faster and address them before attackers can exploit them.
Originally reported by helpnetsecurity.com
Read the original article →Original headline: OpenAI Puts NATO NCIA on Daybreak Cyber Program, Peaking at 300B Tokens a Month