Loading prices...
All news
Glossy royal blue shield with a silver rim and a violet AI chip with gold pins, circled by a cyan orbit ring, near-black navy background with a violet glow and circuit traces

OpenAI says NATO uses up to 300 billion tokens a month for cyber defence

16:30 · 17.11.2026
Source: OpenAI
0

NATO has used up to 300 billion tokens a month of OpenAI's cybersecurity models to protect allied networks, OpenAI said on Tuesday. The figure counts raw input and output tokens. Neither side has disclosed what NATO pays.

The NATO Communications and Information Agency, NCIA, gives the alliance's core institutions access through Daybreak. OpenAI runs the program for vetted defenders doing authorised work: reviewing code, investigating suspicious activity, validating vulnerabilities and testing fixes. In September OpenAI put $1 billion in credits behind it.

OpenAI timed the news to the opening of NATO Edge 26, the agency's industry conference in İzmir.

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

— Dylan Browne, General Manager of NCIA

Quote source: UK Defence Journal, 17 November 2026.

Browne said NCIA will also study token consumption and usage patterns, so it can control costs, compare vendors and plan procurement. He said the models speed up vulnerability discovery and detect new threats at machine speed.

OpenAI launched Daybreak this year. Anthropic runs a similar program for vetted defenders, Project Glasswing, so NCIA can use its cost study to compare the two.

Who else has access

  • ENISA, the EU cyber agency, found four vulnerabilities in software used across EU institutions, and fixes are out for all four
  • CERT Polska helped find six flaws in RouterOS, the system on MikroTik network gear, and confirmed that MikroTik's fixes stop attacks it had seen
  • France's ANSSI and national agencies in Germany and Poland can use the models, and Ukraine's Ministry of Digital Transformation uses Daybreak to defend civilian infrastructure

In a November report ENISA and CERT-EU reviewed 27 software projects with several frontier models, OpenAI's among them. Maintainers assessed 88 potential issues and confirmed 95% of them as real, Help Net Security reported.

Agents in the exercise

At Crossed Swords 2026, which the NATO Cooperative Cyber Defence Centre of Excellence runs in Tallinn, a new training stream had participants build, deploy and test an AI agent for cyber operations inside Daybreak. People stayed in charge as operators.

Ukraine's digital transformation minister Oksana Ferchuk told NATO Edge that AI makes Russian cyberattacks more massive and cheaper, and that defenders cannot rely on human minds anymore.

Attackers get the same capability. In September we wrote that OpenAI's Astra model can build cyberattacks without human help. OpenAI pitches Daybreak as a way to give defenders that speed first.

If you run MikroTik routers, install the latest RouterOS update. CERT Polska saw the six flaws used in real attacks before MikroTik shipped the fixes.

The usage figures come from OpenAI and NCIA, and nobody has checked them independently. This report is for information only.

Published: 16:30 · 17.11.2026
Maks

Author

Maks

Trading man

I've been interested in the cryptocurrency market for a long time, am a trader, and write articles and news about my experience and crypto in simple terms.

Comments (0)

No comments yet — be the first!

Join our community

No spam, unsubscribe in one click