NATO Used Up to 300B OpenAI Tokens a Month for Cyber Defense

NATO has used up to 300 billion input and output tokens a month of OpenAI’s cybersecurity models, OpenAI said in a post dated 17 November. The work runs through NATO’s Communications and Information Agency (NCIA), which provides the alliance’s core institutions with access to the models through OpenAI’s Daybreak programme.

At that peak, 300 billion tokens a month is about 10 billion a day, or about 116,000 tokens every second, by our arithmetic. Neither OpenAI nor NCIA gave a price or contract value in the post.

Business Pill · AI Agents

A short explainer on AI agents, software that plans and carries out multi-step tasks. It teaches the general idea and does not discuss NATO or OpenAI’s cyber models.

The key insight: By our arithmetic, NATO’s peak of up to 300 billion tokens a month is about 116,000 tokens a second, and 60% of the roughly 500 billion a month that Sam Altman cited for the heaviest AI user on 2 November.

What the 300 Billion Tokens Measure

OpenAI wrote that NATO “has used up to 300 billion raw input and output tokens per month as part of this work”. The figure is a monthly peak, not an average, and the post does not say over how many months it was measured.

For scale, OpenAI chief executive Sam Altman said on 2 November that the world’s heaviest AI user consumes about 500 billion tokens a month, as set out in our piece on that remark. NATO’s peak is 60% of that figure, by our arithmetic.

The tools are for authorised defensive work. OpenAI describes Daybreak as putting AI “into the hands of trusted defenders for authorized security work”, from reviewing software to validating vulnerabilities and testing fixes. Its pen-testing agents are covered in our Daybreak launch piece.

AI tokens used per month: NATO’s peak of up to 300 billion through NCIA, per OpenAI’s 17 November
AI tokens used per month: NATO’s peak of up to 300 billion through NCIA, per OpenAI’s 17 November 2026 post, against the roughly 500 billion a month for the heaviest AI user that Sam Altman cited on 2 November. The 60% ratio is our arithmetic.

NATO’s Cost Line

NCIA’s general manager, Dylan Browne, said in the post: “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.”

OpenAI also supported the 2026 Crossed Swords exercise, organised by the NATO Cooperative Cyber Defence Centre of Excellence, in which participants “built, deployed, and evaluated AI agents for cyber operations” using Daybreak, with an emphasis on human oversight, according to the post.

Bar chart: in the ENISA and CERT-EU review cited by OpenAI, 88 potential software issues were assessed by maintainers and about 84 (95%) confirmed as genuine
Potential software issues assessed by maintainers in the November 2026 ENISA and CERT-EU review, as cited in OpenAI’s 17 November post; about 84 confirmed is 95% of 88, our arithmetic.

Europe’s Other Users

OpenAI said it has already given access to its cyber models to the EU’s cyber agency ENISA and to national agencies in France (ANSSI), Germany and Poland, and that Ukraine’s Ministry of Digital Transformation uses Daybreak for civilian infrastructure.

According to the post, ENISA used the models to find four vulnerabilities in software used across EU institutions, with fixes published for all four. It cites a November ENISA and CERT-EU report in which 95% of 88 potential issues assessed by software maintainers were confirmed as genuine, about 84 issues by our arithmetic, across the models tested.

CERT Polska used OpenAI models to help identify six vulnerabilities in MikroTik RouterOS, which MikroTik fixed, the post says. On the attack side, Google’s threat report on AI-run exploits is in our GTIG piece, and OpenAI’s own cyber model ratings in our GPT-6 Cyber piece.

The Structural Read

As we read the post, the disclosed number is volume, not money: OpenAI gives tokens but no spend, and NCIA’s general manager frames the next step as learning the economics in order to compare suppliers.

As we read the list of users, OpenAI names one alliance agency, one EU agency, national agencies in three countries and Ukraine’s digital ministry, all in Europe or working with European defenders.

Dr Dylan Browne, NCIA General Manager, in OpenAI’s post, 17 November 2026

“AI-powered cyber defence models represent a generational leap in cyber security”

Three Implications

FOR THE ATTACK SIDE Google’s report on AI-run exploits is in our GTIG piece.

FOR THE MODELS OpenAI’s cyber model ratings are in our GPT-6 Cyber piece.

FOR THE MONEY Funding for AI security start-ups is in our cyber-funding piece.

The Business Engineer Lens

This story maps onto the Business Engineer framework The CFO’s Guide to the Token Economy.

The framework argues: “In 2026 it became a CFO problem, because the economics finally turned legible — and unforgiving.”

As we read it, NATO’s disclosure fits that shift: the alliance’s IT agency is already counting tokens by the hundred billion and says it wants the economics in order to optimise costs and compare suppliers. The framework predates the NATO post and does not discuss it.

What Is Not Established

The post does not give NATO’s spend, the contract terms, which models were used, or how the 300 billion tokens split between input and output, so no cost can be derived from it. “Up to” means the monthly figure is a ceiling reached at least once, not a typical month.

The comparison with the 500 billion figure sets a company post against a spoken remark from a different event. We did not contact OpenAI or NCIA for this piece.

This Week In The AI Supercycle

AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source

This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.

Read the full issue →

Business Engineer Framework

The CFO’s Guide to the Token Economy

The Business Engineer premium analysis of why cheaper tokens produce bigger AI bills.

Read the premium analysis →Get 50% off →

The Bottom Line

OpenAI says NATO has used up to 300 billion tokens a month of its cyber models through NCIA, about 116,000 tokens a second at that peak by our arithmetic, and NCIA’s head says that as adoption scales, the agency will build an understanding of the economics to help evaluate competing solutions. Funding for AI security start-ups is in our cyber-funding piece.

95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. The NATO figures, the NCIA quote and the details on ENISA, CERT Polska, Ukraine and the Crossed Swords exercise are from OpenAI’s own post, dated 17 November 2026 in its news feed, read in full; the ENISA and CERT-EU review figures are as OpenAI cites them. The 500 billion figure is Sam Altman’s remark at ADIPEC on 2 November, from our earlier piece. Daily and per-second rates and percentages are our arithmetic. OpenAI and NCIA were not contacted for this piece.

Sources: OpenAI, Strengthening NATO and allied cyber defenses with AI (17 Nov 2026) · CCDCOE, Exercise Crossed Swords 2026 (as linked by OpenAI) · ENISA and CERT-EU joint report, November 2026 (as cited by OpenAI)

Scroll to Top
FourWeekMBA