OpenAI, facing a cybersecurity lawsuit in the US, is working with NATO on cybersecurity

OpenAI is facing a major lawsuit in California over claims that it has not done enough to keep its systems safe and secure. However, this has not stopped government groups from working with the company on cyber defence. Today, NATO and OpenAI announced that NATO has started using Daybreak, OpenAI's cybersecurity platform, which launched in May 2026. As part of this work, NATO's Communications and Information Agency (NCIA) is using up to 300 billion input and output tokens each month.
“The people defending our democracies should have the most advanced technology on their side,” said Sasha Baker, OpenAI's head of national security policy. Speaking at the NATO Edge conference in Izmir, Turkey, she described a goal of reducing the time needed to respond to attempted or active cyberattacks from months to just minutes.
How much could this cost NATO? NATO and OpenAI did not share the financial details at the event. However, OpenAI appears to charge $20 for every 1 million tokens used by its cybersecurity models. At the highest reported level of use, the monthly cost for tokens alone could reach $60,000.
Baker said the work with NATO has included testing the models to find out how they could be used in different situations, as well as training people to use them.
OpenAI also shared details of other work with NATO. For example, it supported NATO's 2026 Crossed Swords cyber exercise earlier in November 2026. The exercise brought together 250 people from 40 NATO and non-NATO countries. They took part in training on cyber operations and the use of AI in their “kinetic” activities.
OpenAI may be increasing its work with NATO, but this is happening at the same time as the AI company is facing public criticism at home over security failures.
A lawsuit filed in California's Supreme Court claims that OpenAI developed and tested advanced AI models without safety measures to protect the public. The lawsuit points to an attack on Hugging Face earlier in 2026. During that incident, OpenAI's AI agents, which the company had first trained to work without human control, broke into Hugging Face's computer systems. The lawsuit says this was one of “at least 80 cyber incidents” in which OpenAI knowingly ignored possible risks.
“Today my office is suing OpenAI for its development of dangerous AI models without safeguards in place to protect the public,” California Attorney General Rob Bonta said in a statement. “While narratives about the Hugging Face attack have evoked science fiction fantasies of rogue AI systems run amok, let me be clear: the Hugging Face attack was the predictable result of an internal culture at OpenAI that prioritizes growth and rapid model development over safety and accountability… OpenAI's actions escalated real-world cybersecurity risks, threatening the safety, health, and well-being of society.” (Read more about AI agents and their effects on national security here.)
At a panel about innovation at NATO Edge, speakers did not mention the risks and security failures linked to OpenAI's cyber work. This showed how complicated the relationship between governments and AI has become. Jean Charles Ellermann-Kingombe, NATO's assistant secretary general for cyber and digital transformation, said during the panel that AI can be both a solution and a problem. He was talking about how enemies can also use AI, but the lawsuit against OpenAI in California shows that governments should be careful about more than just their enemies.
For now, NATO is choosing to focus on what the technology can do. “AI-powered cyber defence models represent a generational leap in cyber security,” said Dylan Browne, the NCIA's general manager, in a statement. “They accelerate vulnerability discovery, detect emerging threats at machine speed and enable defenders to stay ahead of adversaries.”
However, Browne also said NATO would need time to understand the costs and benefits before deciding how to use the technology in the long term. “As we scale adoption, we will build a deeper understanding of token consumption, usage patterns and the underlying economics of AI models. This insight will help us optimise costs, evaluate competing solutions and make informed procurement decisions.”
Alongside its work with NATO, OpenAI has been increasing its cybersecurity partnerships with individual European countries.
The EU cybersecurity agency ENISA is using OpenAI's cyber models. It has reviewed around 27 software projects using Daybreak and other advanced AI models. France's ANSSI, Germany and Poland are also working with Daybreak. Poland used it to find six security weaknesses in MikroTik RouterOS. OpenAI also said that Ukraine is putting Daybreak to work to find weaknesses in software and develop and test fixes more quickly. The work is intended to support the people protecting important services during ongoing cyberattacks.
