NATO's AI cyber defense work hits up to 300 billion OpenAI tokens a month
NATO's cybersecurity efforts utilize OpenAI's models, processing up to 300 billion tokens monthly to enhance network protection for Allied nations.

NATO’s Communications and Information Agency (NCIA) reports that it processes up to 300 billion raw input and output tokens each month through OpenAI’s cybersecurity models, aimed at protecting Allied networks. This access is facilitated via Daybreak, a program that allows vetted defenders to perform authorized tasks such as code reviews and threat assessments.
NATO’s NCIA has integrated OpenAI's cybersecurity capabilities to bolster the security of its member nations' networks. The substantial volume of up to 300 billion tokens processed monthly indicates a significant reliance on AI technology for cybersecurity efforts. This collaboration aims to enhance the efficiency and effectiveness of threat detection and response across Allied networks, allowing for quicker identification and remediation of vulnerabilities.
The process works by utilizing OpenAI's models through the Daybreak platform, which is designed for vetted defenders. These defenders can review code, investigate suspicious activities, and test potential fixes. This structured approach not only enables the identification of vulnerabilities but also streamlines the implementation of necessary security measures, thereby improving overall network resilience.
AI models, like those from OpenAI, operate by analyzing vast amounts of data to identify patterns and anomalies. In the context of cybersecurity, these models can sift through code and network traffic to detect vulnerabilities or unusual behavior that may indicate a cyber threat. By leveraging machine learning, these systems can continuously improve their detection capabilities as they process more data, making them valuable tools for security teams.
Various entities are benefiting from this initiative, including the European Union’s ENISA and CERT-EU. These organizations have employed OpenAI’s models to identify vulnerabilities in software used throughout EU institutions, successfully identifying four vulnerabilities that have since been addressed. Additionally, Ukraine’s Ministry of Digital Transformation and national agencies in France, Germany, and Poland are also utilizing these AI capabilities to enhance their cybersecurity postures.
NATO’s collaboration with OpenAI has been showcased during the 2026 Crossed Swords exercise, where participants developed and tested AI agents for cyber operations.
Dr. Dylan Browne, NCIA’s general manager, noted that as adoption scales, NATO aims to better understand token consumption and usage patterns, which will inform procurement decisions and cost optimization. This focus on data-driven insights will help NATO refine its cybersecurity strategies as it continues to integrate AI technologies.
Currently, it remains unclear how these AI models will evolve in response to emerging threats and whether additional vulnerabilities may be discovered as more organizations adopt similar technologies. Observers will be watching for updates on the effectiveness of these models in real-world scenarios and any new findings that may arise from ongoing AI-assisted cybersecurity efforts.
What it meansNATO is leveraging OpenAI's AI models to enhance cybersecurity for its member nations, processing a large volume of data monthly.
Ask your IT team this weekWhich AI models are we currently using for cybersecurity?How many vulnerabilities have we identified using AI tools?
DecisionThe IT team will evaluate the effectiveness of current AI models in detecting vulnerabilities, and security will check the results.