TechAILatest.newsTrackers
Cybersecurity / AI security tools

OpenAI gives NATO its Daybreak cyber models; alliance used up to 300 billion tokens a month

OpenAI said on 17 November 2026 that NATO's NCIA uses its Daybreak cyber defense models, at up to 300 billion tokens a month, alongside EU and national agencies.

By Emre A.Published

Abstract illustration of a shield over a networkTech AI Latest
Illustration of a shield over a network, made by Tech AI Latest for this story. Illustration: Tech AI Latest (Original illustration)

OpenAI announced on 17 November 2026 that NATO's Communications and Information Agency (NCIA) has access to its cybersecurity models through the Daybreak program to protect allied networks, and said the alliance has used up to 300 billion input and output tokens a month. EU cyber agency ENISA, national agencies in France, Germany and Poland, and Ukraine's digital ministry also use the models. No price or contract value was published.

Key facts

  • Announced: 17 November 2026 (OpenAI)
  • Customer: NATO Communications and Information Agency (NCIA), for the alliance's core institutions
  • Usage: up to 300 billion raw input and output tokens a month, according to OpenAI
  • Other users: ENISA; national cyber agencies in France, Germany and Poland; Ukraine's Ministry of Digital Transformation
  • Exercise: Crossed Swords 2026, run by NATO's cyber defence centre in Tallinn, included a stream where teams built AI agents with Daybreak
  • Not stated: price, contract value, duration and which models NATO uses

What OpenAI announced

OpenAI said on 17 November 2026, in a post titled Strengthening NATO and allied cyber defences with AI, that the NATO Communications and Information Agency (NCIA) now gives the alliance's core institutions access to its advanced cybersecurity models. The models are provided through Daybreak, OpenAI's program for governed access to its most capable security tools, and are meant to protect allied networks.

OpenAI says NATO has already used up to about 300 billion raw input and output tokens a month. That is a peak figure and OpenAI's own claim; neither NATO nor NCIA has published usage numbers, and coverage by Quartz and Anadolu Agency attributes it to the company. Reported uses include reviewing code, scanning for unusual network behaviour and confirming weaknesses before they are fixed.

The post also describes Crossed Swords 2026, the exercise run by the NATO Cooperative Cyber Defence Centre of Excellence in Tallinn. It included a training stream in which participants built, deployed and evaluated AI agents for cyber operations using Daybreak. OpenAI did not publish a price, a contract value, a duration or the names of the specific models NATO uses.

The European agencies in the program

NATO is not the only public body named. OpenAI says the European Union Agency for Cybersecurity (ENISA) and national cyber agencies in France, Germany and Poland are working with Daybreak, and that Ukraine's Ministry of Digital Transformation uses it to strengthen the defence of civilian infrastructure, according to Benzinga and Quartz.

Poland offers the clearest public example of what government use looks like. In September, CERT Polska, the national incident response team, disclosed six vulnerabilities in MikroTik RouterOS that attackers were already chaining to take over routers with SSH exposed to the internet. CERT Polska said the flaws were found with an AI-assisted research environment using OpenAI's GPT-5.5-cyber and GPT-5.6-sol models. MikroTik shipped fixes in RouterOS 7.25beta3, 7.24.2, 7.23.4 and 6.49.21.

The NATO deal extends a month of Daybreak expansion. OpenAI released GPT-6 Cyber for approved Daybreak Red customers on 5 November, rated Critical under its Preparedness Framework, and launched Daybreak Red Team and Defense Factory on 12 November.

Why it matters

Governments have so far been cautious about putting frontier AI models near defence networks. In March 2026, reports said OpenAI was in talks to deploy on NATO's unclassified networks, and a company spokesperson later stressed that the opportunity covered unclassified networks only. The 17 November post does not say which NATO networks the models touch, so that boundary remains unclear from the public record.

The scale claim matters for a second reason. Tokens are how AI vendors bill, and a peak of 300 billion a month suggests NATO is running automated analysis continuously rather than as occasional chat sessions. Without a published price it is impossible to say what that costs or how it compares with offers from rivals. Anthropic, Google and Microsoft run their own AI vulnerability programs, compared in this site's guide to AI bug hunters.

The timing is awkward for OpenAI. A day earlier, California sued the company over the containment of its test agents, citing the Hugging Face breach. Allied agencies now rely on the same company's models for defence.

What to watch next

Open questions include whether NCIA publishes its own account of results, whether other NATO members' national agencies join, and whether the access extends beyond unclassified systems. Pricing, the length of the agreement and how findings are disclosed to software vendors are also not public. More cybersecurity coverage on this site tracks Daybreak and rival programs as they expand.

What changed

BeforeDaybreak access for approved companies and some national agencies; NATO talks reported in March 2026
AfterNATO's NCIA has Daybreak access for core alliance institutions and used up to 300 billion tokens a month, per OpenAI
Who it affectsNATO and allied cyber agencies, EU agencies such as ENISA, vendors whose software these agencies review

Questions people ask

What did OpenAI agree with NATO?
OpenAI says NATO's Communications and Information Agency gives the alliance's core institutions access to its cybersecurity models through the Daybreak program, to protect allied networks. It announced this on 17 November 2026.
How much does NATO use OpenAI's models?
OpenAI says NATO has used up to about 300 billion raw input and output tokens a month. NATO has not published its own figure, and no price or contract value was disclosed.
Which other governments use OpenAI Daybreak?
OpenAI names ENISA, national cyber agencies in France, Germany and Poland, and Ukraine's Ministry of Digital Transformation.

Sources

  1. Strengthening NATO and allied cyber defences with AI (OpenAI, 17 Nov 2026)
  2. OpenAI gives NATO access to Daybreak cybersecurity AI models (Quartz, 17 Nov 2026)
  3. OpenAI makes advanced cyber defense models available to NATO (Anadolu Agency, 17 Nov 2026)
  4. OpenAI Expands AI Cybersecurity Push Across NATO, European Governments (Benzinga, 17 Nov 2026)
  5. Vulnerabilities in MikroTik RouterOS actively exploited (CERT Polska, 5 Sep 2026)

Reference pages

Spotted an error? See the corrections policy and tell the editor. Facts in this story come from the sources listed above; vendor claims are attributed.

Comments

Loading comments.

No links, no insults; comments that break the rules are not published. 0/1000

Related stories

All cybersecurity