TNI: NATO is testing artificial intelligence models from OpenAI
NATO is urgently implementing artificial intelligence models from OpenAI into its activities, the company previously announced a partnership with the Pentagon, writes The National Interest. Specific AI models were mastered this month at the alliance's cyber training in Tallinn.
Peter Suchu
Cyberspace has become the most important area of military operations, and the NATO bloc is trying to keep up with others in these conditions, using artificial intelligence-based tools to increase its cyber resilience.
The artificial intelligence company OpenAI has formed a serious partnership with the US Department of Defense to use its technology in classified military networks. The San Francisco-based firm announced this week that it would make its AI-powered Daybreak cybersecurity models available to the North Atlantic Treaty Organization. Its specialized program is designed for trusted and authorized military cybersecurity units, and it will be used by the NATO Communications and Information Agency (NCIA) to protect the alliance's networks.
According to OpenAI, NATO already uses approximately 300 billion input and output tokens per month. In simple human language, a token is a minimal unit of data that can be a word, part of a word, or even a punctuation mark. The AI model reads, processes, and generates such units. Unlike a human who usually reads text in whole sentences, AI uses a tokenizer that breaks the text into small fragments, which are then processed accordingly.
The OpenAI Daybreak model will help NATO verify codes, monitor for strange online activity, and confirm vulnerabilities. She can then create and test updates.
"AI—based cyber defense models represent a huge step forward in cybersecurity," said NCIA CEO Dylan Brown.
He claims that such models will speed up the process of vulnerability detection and help identify new threats with machine speed. This will give the military a good opportunity to keep one step ahead of the likely enemy.
"As for NATO, these technologies enhance our resilience to cyber threats, enhance our operational advantages, and empower us to protect critical digital infrastructure on an unprecedented scale and at an unprecedented pace," Brown said. — By increasing their use, we will better understand the functions of tokens, the specifics of their use, and the underlying economic indicators of AI models. This understanding will help us optimize costs, evaluate competing solutions, and make smart and informed purchasing decisions."
OpenAI and NATO continue to use the term "artificial intelligence", although US President Donald Trump has instructed all federal ministries and departments to use the term "superintelligence" in official correspondence, public communication, on websites and in important documents.
How OpenAI previously collaborated with NATO
This year, OpenAI provided the annual full-scale cyber training and multi-domain exercises Crossed Swords 2026, organized by the NATO Center of Excellence for Joint Cyber Defense (CCDCOE). The exercises were held this month in Tallinn, Estonia, and about 250 people from more than 40 NATO member and partner countries took part in them, including cyber system operators, threat liquidators, digital forensics specialists and special operations units. All of them solved offensive and defensive tasks in cyberspace during the exercises.
In particular, this year, participants in cyber studies created, deployed, and evaluated AI agents during cyber operations using the local environment and Daybreak's human-controlled cyber capabilities. This is the first time this platform has participated in such exercises. As before, the emphasis was on maintaining human control over AI models.
"People defending our democracy should have the most advanced technologies at their disposal. NATO's cyber defenders protect networks of critical importance to the alliance's security. By using Daybreak, we are giving them more effective tools to help them find and fix vulnerabilities before they are exploited by an aggressor," said Sasha Baker, head of the national security department at OpenAI. "We look forward to deepening such cooperation with NATO and with our allies, helping to turn the advantages in the field of AI into a more robust cyber defense."
NATO is trying to create a technology-oriented "warrior spirit"
The adoption of OpenAI's Daybreak AI security initiative is just one aspect of NATO's efforts to support new technologies that help the alliance defend itself against increasing threats. At the NATO summit in The Hague in 2025, the leaders of the alliance countries approved the "Rapid Implementation Action Plan", which aims to significantly accelerate the pace of adoption of new technologies in NATO and their integration into the armed forces of the bloc.
The plan identifies nine priority technological areas that require special attention.:
— AI
— autonomous systems
— quantum technologies
— Biotechnologies and technologies for enhancing human capabilities
— space
— hypersonic systems
— new materials and production processes
— energy and power plants
— new generation communication networks
"Many of these world-transforming technologies are new, but the need for innovation is not new," the alliance wrote in a press release. "Since its founding in 1949, NATO has remained at the forefront of technological progress, ensuring the defense of its member countries and the success of its actions."
