OpenAI Daybreak Helps NATO Defend Allied Networks With AI

0

OpenAI says NATO's technology agency now has access to its advanced cybersecurity models, which are meant to help defenders find and fix software weaknesses.

OpenAI and NATO cybersecurity partnership represented by a digital security shield and handshake.
OpenAI and NATO partner to strengthen cybersecurity defenses using artificial intelligence.

NATO's core institutions can now use OpenAI's advanced AI cybersecurity models to defend allied networks. OpenAI announced the work on November 17, 2026, saying it is collaborating with the NATO Communications and Information Agency (NCIA) through its Daybreak cybersecurity program.

What OpenAI Says NATO Is Using the Models For

According to OpenAI, the models are intended for authorized defensive work only. The company lists four main uses:

  • Reviewing software
  • Investigating suspicious activity
  • Validating vulnerabilities
  • Developing and testing security fixes

In practice, that means the models help security teams look for weaknesses in code, check whether a suspected flaw is real, and work on patches before attackers can exploit the problem.

A Large Usage Figure, Reported by the Company

OpenAI reports that NATO has used up to 300 billion raw input and output tokens per month. Tokens are the small units of text that AI models read and generate, so the number points to heavy usage. The figure comes from OpenAI. It has not been independently audited, and it does not by itself show how effective the tools have been.

Crossed Swords 2026 and the Role of Human Oversight

OpenAI also says it supported the 2026 Crossed Swords exercise. NATO's Cooperative Cyber Defence Centre of Excellence (CCDCOE) confirms the exercise ran from November 2 to November 6 in Tallinn, Estonia. It drew more than 250 participants from over 40 countries.

The CCDCOE says one training stream focused on designing, developing, deploying and evaluating AI agents for cyber operations. Human oversight was an important part of the exercise, which matters because the stated purpose of the Daybreak work is to support defenders rather than replace them.

Other European Agencies Are Also Involved

NATO is not the only organization named. OpenAI says it has given access to its cybersecurity models to European defenders, including the EU Agency for Cybersecurity (ENISA) and national cybersecurity agencies in France, Germany and Poland. It also says Ukraine's Ministry of Digital Transformation is using Daybreak to help protect civilian infrastructure.

OpenAI points to several results from these partners:

  • ENISA: The agency used the models to identify four vulnerabilities in software used across EU institutions, and fixes were published for all four.
  • ENISA and CERT-EU: A joint report this month examined 27 software projects. Across the models tested, maintainers confirmed 95% of 88 assessed potential issues as genuine.
  • CERT Polska: The team used OpenAI models to help identify six vulnerabilities in MikroTik RouterOS. MikroTik released fixes, which CERT Polska said prevented the attacks it had observed.

These results are as reported by OpenAI. The underlying ENISA and CERT Polska records have not been independently reviewed for this report.

Background: A Possible Contract Was Reported in March

On March 3, 2026, Reuters reported that OpenAI was considering a contract with NATO for unclassified networks. That report was about a possible deal. It does not establish that the current work began in March, and November 17 is the date of OpenAI's announcement, not a confirmed start date.

What Remains Unclear

Several details have not been made public in the information available so far:

  • When the collaboration actually began
  • Any financial value or contract terms
  • Whether the models are used on classified networks
  • Any independently measured improvement in cybersecurity

OpenAI describes the models as tools for authorized defenders. The announcement does not say that AI has replaced NATO's cybersecurity professionals, and it does not claim the arrangement guarantees protection against cyberattacks.

Latest Confirmed Update

As of November 17, 2026, the confirmed facts are that OpenAI has announced its work with NATO's NCIA through Daybreak, and that NATO's CCDCOE has confirmed Crossed Swords 2026 and its AI-agent training stream. The usage figures and the ENISA and CERT Polska findings rest on OpenAI's own account for now, so independent confirmation from NATO or the agencies involved would be the next thing to watch for.

Post a Comment

0 Comments
* Please Don't Spam Here. All the Comments are Reviewed by Admin.
Post a Comment (0)
To Top