Updated
Updated · OpenAI · Nov 17
OpenAI Expands AI Cyber Defense to NATO, Allies as NATO Uses 300 Billion Tokens Monthly
Updated
Updated · OpenAI · Nov 17

OpenAI Expands AI Cyber Defense to NATO, Allies as NATO Uses 300 Billion Tokens Monthly

3 articles · Updated · OpenAI · Nov 17

Summary

  • NATO’s Communications and Information Agency now uses OpenAI cybersecurity models across core institutions, with usage reaching up to 300 billion raw input and output tokens a month to help protect Allied networks.
  • OpenAI said the tools support authorized defense work including software review, suspicious-activity investigations, vulnerability validation, and developing and testing fixes through its Daybreak program.
  • Ukraine’s digital ministry is using Daybreak to defend civilian infrastructure, while ENISA found four vulnerabilities in software used across EU institutions and reviewed 27 software projects with frontier models.
  • CERT Polska used OpenAI models to identify six MikroTik RouterOS vulnerabilities, and MikroTik issued fixes that the agency said blocked the attacks it had observed.
  • The rollout also fed NATO’s 2026 Crossed Swords exercise, where participants built and tested AI agents for cyber operations under human oversight, underscoring broader allied adoption of AI in defense.

Insights

How will OpenAI prevent adversaries from weaponizing the same frontier AI models now defending NATO and EU networks?
What happens when state-backed hackers successfully extract the advanced cyber capabilities embedded within OpenAI's defensive models?
If AI drastically shrinks the time from vulnerability discovery to exploitation, can human defenders ever patch systems fast enough?