OpenAI grants NATO access to advanced cybersecurity AI models
NATO governments will use OpenAI's most powerful models to identify and patch software vulnerabilities as cyber threats from Russia intensify.

The short version
- NATO governments gain access to OpenAI's most advanced models to find and patch software vulnerabilities.
- The move follows a cyber wargame in Estonia where operators deployed AI agents to complete missions.
- The NATO Communications and Information Agency plans to award over $3 billion in contracts for AI and cloud services by 2027.
- Experts warn that AI allows hostile actors to create more effective malware and find software flaws faster.
Why it matters
This partnership signals a shift toward 'agentic' defense, where AI actively works alongside humans to secure infrastructure. It aims to move the speed of cyber defense from weeks to minutes to counter state-sponsored threats.
A new era of cyber defense
NATO governments are set to receive access to OpenAI's most powerful artificial intelligence models. This partnership is designed specifically to help the alliance find and patch vulnerabilities in critical software. The announcement was made during NATO Edge, a major technology event held in Izmir, Turkey.
The move comes at a time when Russia is increasing its cyberattacks on NATO allies. Security officials have long warned that these types of attacks are becoming more sophisticated. Experts believe that AI will allow hostile actors to multiply their efforts and create more effective malware.
By providing these tools, OpenAI aims to help the alliance stay ahead of these threats. The goal is to create a defensive posture that can keep up with the speed of modern cyber warfare. This collaboration marks a significant expansion of OpenAI's efforts to provide high-level security tools to government entities.
Testing AI in the field
The alliance recently tested these capabilities during a cyber wargame called Crossed Swords in Tallinn, Estonia. During the exercise, NATO operators designed and deployed AI agents to achieve specific missions. These agents functioned as tools that the human operators worked alongside to complete their objectives.
The technical environment for the exercise was complex and utilized various types of technology. It included on-premises AI systems and, for the first time, OpenAI Daybreak. Daybreak is a specific initiative that provides accredited institutions and the U.S. government with access to advanced models for vulnerability research.
This practical application of AI agents is part of a broader strategy to modernize defense. Instead of just using AI as a passive tool, the alliance is exploring agentic AI. This means the AI can perform multi-step tasks to find and fix security flaws automatically.
The Daybreak initiative
OpenAI Daybreak was originally launched in May to support high-level security needs. It provides a way for the U.S. government and other accredited institutions to use the company's most advanced models. The primary focus of this program is identifying weaknesses in key pieces of software before they can be exploited.
Anthropic, another major AI company, has launched a similar project called Project Glasswing. Both companies are moving toward a philosophy where AI is used to find and fix vulnerabilities first. This proactive approach is intended to neutralize threats before they can cause significant damage to infrastructure.
By integrating Daybreak into NATO's toolkit, OpenAI is expanding its reach into international security. The goal is to provide a standardized way for allies to protect their digital borders. This ensures that the most advanced technology is available to those defending against large-scale cyber threats.
Investment and infrastructure
The NATO Communications and Information Agency, or NCIA, is making a massive investment in new technologies. The agency plans to award more than $3 billion in contracts to the private sector by the end of 2027. These contracts will focus heavily on cloud computing, space services, and artificial intelligence.
This $3 billion figure represents the specific funding from the agency. It does not include the additional amounts that individual NATO allies spend on their own AI initiatives. This indicates a significant push to modernize the alliance's underlying technological framework.
Many exhibitors at the NATO Edge event offered AI-powered products to meet these growing demands. The alliance is looking for ways to integrate these tools into their existing defense commitments. This investment is intended to create a more resilient and technologically advanced security network for all members.
Shrinking the time window
The ultimate goal of this partnership is to change the speed of cybersecurity responses. Sasha Baker, OpenAI's national security and policy head, emphasized the need to reduce the time between discovery and patching. She wants to move the timeline from months to weeks, and eventually to hours or even minutes.
This shift is necessary because attackers can exploit a flaw almost as soon as it is discovered. If defenders take weeks to react, the damage can be widespread and severe. Using AI agents allows for a much faster response that can mitigate risks in real-time.
OpenAI is encouraging the alliance to hold them accountable for achieving these speed benchmarks. The success of this initiative will be measured by how quickly the alliance can secure its systems. By automating the discovery and patching process, NATO hopes to protect Western populations from disruption and government frustration.
What to watch
- How quickly NATO can successfully reduce patch times from weeks to minutes.
- Whether other AI companies will launch similar 'Daybreak' style programs for other international alliances.
Try it
Look up the 'Crossed Swords' cyber wargame to see how NATO is currently testing AI agents in real-world scenarios.
Sources and links
- OfficialProject Glasswing anthropic.com
- Officialhelp.openai.com/en/articles/20001258-openai-daybreak-trusted-access-fo help.openai.com
- ReportingDefense One: the article we sourced
First reported by Defense One. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.