Rudra Kasturi Inc.

Article · LLMs

OpenAI gives NATO access to AI cybersecurity models: what the Daybreak work involves

SUMMARY

OpenAI says NATO's Communications and Information Agency is using its cybersecurity models for authorised defensive work, including vulnerability investigation and remediation.

OpenAI says NATO’s Communications and Information Agency (NCIA) is working with its advanced cybersecurity models to help protect allied networks. The announcement, published on 17 November 2026, describes authorised defensive uses such as reviewing software, investigating suspicious activity, validating vulnerabilities and developing and testing fixes.

The work is part of OpenAI’s Daybreak cybersecurity programme. The company says NATO has used up to 300 billion raw input and output tokens per month through this work. That figure is a company-reported usage measure, not an independent measure of cyber incidents prevented or vulnerabilities fixed.

What is the AI being used for?

Cybersecurity teams work across large volumes of code, alerts, system behaviour and technical documentation. AI models can help defenders examine software, investigate suspicious activity and explore possible weaknesses more quickly.

OpenAI also says the 2026 Crossed Swords exercise included a training stream in which participants built, deployed and evaluated AI agents for cyber operations. The company emphasises that human operators remained involved.

The distinction matters. Defensive AI can help authorised teams find and fix weaknesses, but the same broad capabilities can be misused. Access controls, oversight and clear authorisation are essential.

Why NATO’s use is significant

NATO’s institutions and member countries depend on complex networks and digital infrastructure. A vulnerability in software used by a government or critical service can have consequences beyond one organisation.

AI-assisted analysis could help security teams work through large codebases and investigate issues faster. OpenAI says its work also extends to allied defenders, including Ukraine’s Ministry of Digital Transformation and European national cyber agencies.

The company reports that the EU cybersecurity agency ENISA used its models to identify four vulnerabilities in software used by EU institutions, with fixes published for all four. That is a company account of the programme and should be read alongside the underlying technical disclosures.

What organisations can learn

  1. Use AI tools only within clearly authorised security environments.
  2. Keep human experts responsible for validating findings and approving fixes.
  3. Test suggested patches before deployment.
  4. Log model access, tool calls and changes to systems.
  5. Measure confirmed vulnerabilities and remediation time, not token usage alone.

The wider direction

Cybersecurity is becoming one of the most important enterprise uses of advanced AI. The opportunity is faster analysis and remediation. The risk is giving powerful tools broad access without adequate controls.

Bottom line: OpenAI’s NATO work shows how frontier AI is being applied to defensive cybersecurity. The important measure will be verified security outcomes, with human oversight kept firmly in the process.

Sources