OpenAI is taking its dedicated cyber tools to one of the world's most attacked infrastructures. The company will give Ukrainian professionals access to the Daybreak program to speed up the search for vulnerabilities and protect civilian critical infrastructure.
OpenAI has signed an agreement with the Ministry of Digital Transformation of Ukraine. The proven teams responsible for critical systems will receive the company's cyber models, training and technical support. AI plans to apply to find weaknesses in networks, check the real exploitability of vulnerabilities, prepare fixes and analyze huge flows of event logs.
Daybreak is an OpenAI trusted access program for cybersecurity models. The company divided the program into several levels. Day Breakt Blue is designed primarily for protective tasks, including malware analysis, incident response, vulnerability search, and fix checks. Daybreak Red is designed for more complex authorized work, including penetration testing and exploit checking.
It is not just about issuing Ukrainian specialists of ordinary ChatGPT with weakened filters. As part of the Daybreak, OpenAI has assembled models and tools focused on practical cybersecurity. Access is given after verification, and the work remains limited to the permitted systems. The company separately prohibits the use of the program to attack foreign infrastructure.
For Ukraine, the main advantage should be speed. The representative of the country Dmitry Kushneruk said that CERT-UA registers about 6000 serious cyber attacks a year, that is, about 15 per day. People have to deal with thousands of journals and warnings, while AI is able to sort events faster, look for really dangerous activity and help specialists focus on priority threats. In early 2026, CERT-UA, for example, recorded a mass campaign with letters on behalf of Ukrainian state bodies.
Ukrainian infrastructure has already experienced attacks with large-scale consequences. In December 2023, the Kyivstar hack disrupted mobile and internet access for millions of subscribers, and attackers gained deep access to the operator's infrastructure. The consequences of the attack affected thousands of servers and communication services.
OpenAI considers Ukraine simultaneously as a recipient of technology and an environment where the capabilities of protective AI can be tested at constant pressure on real systems. Kushneruk called cooperation bilateral. OpenAI provides advanced models and training, and Ukrainian teams give experience of responding to attacks in conditions that are difficult to reproduce in the laboratory.
The company links the project with the concept of “protector’s window”. The point of the approach is to pass on powerful AI tools to critical infrastructure operators as quickly as possible, until similar capabilities become widespread among the attackers. The OpenAI support program has previously allocated the equivalent of $1 billion in subsidized tokens.
The OpenAI partnership announcement highlights the civic focus of the project. Ukrainian teams should use models to protect vital services, find vulnerabilities, and reduce the time between detecting and fixing the problem.
The potential of such systems is already far beyond the analysis of journals. During recent tests, OpenAI AI agents infiltrated the Hugging Face infrastructure, found vulnerabilities and reached production systems. The experiment simultaneously demonstrated the capabilities of autonomous agents and the risks that arise when providing models with broad powers.
Even further advanced is OpenAI’s specialized cyber model. In the tests, the new system hacked the browser, discovered zero-day vulnerabilities and built a chain of privilege increases to root. It is the growth of such opportunities that forces the company to accelerate the transfer of powerful tools to defenders and at the same time strengthen access control.
OpenAI has signed an agreement with the Ministry of Digital Transformation of Ukraine. The proven teams responsible for critical systems will receive the company's cyber models, training and technical support. AI plans to apply to find weaknesses in networks, check the real exploitability of vulnerabilities, prepare fixes and analyze huge flows of event logs.
Daybreak is an OpenAI trusted access program for cybersecurity models. The company divided the program into several levels. Day Breakt Blue is designed primarily for protective tasks, including malware analysis, incident response, vulnerability search, and fix checks. Daybreak Red is designed for more complex authorized work, including penetration testing and exploit checking.
It is not just about issuing Ukrainian specialists of ordinary ChatGPT with weakened filters. As part of the Daybreak, OpenAI has assembled models and tools focused on practical cybersecurity. Access is given after verification, and the work remains limited to the permitted systems. The company separately prohibits the use of the program to attack foreign infrastructure.
For Ukraine, the main advantage should be speed. The representative of the country Dmitry Kushneruk said that CERT-UA registers about 6000 serious cyber attacks a year, that is, about 15 per day. People have to deal with thousands of journals and warnings, while AI is able to sort events faster, look for really dangerous activity and help specialists focus on priority threats. In early 2026, CERT-UA, for example, recorded a mass campaign with letters on behalf of Ukrainian state bodies.
Ukrainian infrastructure has already experienced attacks with large-scale consequences. In December 2023, the Kyivstar hack disrupted mobile and internet access for millions of subscribers, and attackers gained deep access to the operator's infrastructure. The consequences of the attack affected thousands of servers and communication services.
OpenAI considers Ukraine simultaneously as a recipient of technology and an environment where the capabilities of protective AI can be tested at constant pressure on real systems. Kushneruk called cooperation bilateral. OpenAI provides advanced models and training, and Ukrainian teams give experience of responding to attacks in conditions that are difficult to reproduce in the laboratory.
The company links the project with the concept of “protector’s window”. The point of the approach is to pass on powerful AI tools to critical infrastructure operators as quickly as possible, until similar capabilities become widespread among the attackers. The OpenAI support program has previously allocated the equivalent of $1 billion in subsidized tokens.
The OpenAI partnership announcement highlights the civic focus of the project. Ukrainian teams should use models to protect vital services, find vulnerabilities, and reduce the time between detecting and fixing the problem.
The potential of such systems is already far beyond the analysis of journals. During recent tests, OpenAI AI agents infiltrated the Hugging Face infrastructure, found vulnerabilities and reached production systems. The experiment simultaneously demonstrated the capabilities of autonomous agents and the risks that arise when providing models with broad powers.
Even further advanced is OpenAI’s specialized cyber model. In the tests, the new system hacked the browser, discovered zero-day vulnerabilities and built a chain of privilege increases to root. It is the growth of such opportunities that forces the company to accelerate the transfer of powerful tools to defenders and at the same time strengthen access control.
