OpenAI has partnered with Poland’s national research institute NASK to advance artificial intelligence safety, cybersecurity, and technical research. The agreement will give Polish researchers access to advanced AI models and testing methods to evaluate emerging risks, identify vulnerabilities, and strengthen safeguards.
The collaboration will be led by NASK’s AI Safety Research Center, which serves as Poland’s Artificial Intelligence Safety Institute (AISI). The initiative aims to improve the evaluation of increasingly capable AI systems while strengthening international cooperation between public research institutions and AI developers.
AI Models Help Identify Critical Security Vulnerabilities
The partnership builds on earlier cybersecurity research conducted by CERT Polska, NASK’s incident response team. In July, researchers gained access to OpenAI’s cyber-capable models through a program called Daybreak and used them to examine MikroTik RouterOS, an operating system used in networking equipment.
The investigation uncovered six vulnerabilities, including flaws that could be combined to take control of affected devices without authentication. MikroTik subsequently released security patches, and CERT Polska verified that the fixes prevented the attacks identified during testing.
The findings demonstrate how advanced AI models can support vulnerability discovery and accelerate security remediation. By combining automated analysis with human expertise and vendor collaboration, researchers can identify weaknesses and validate whether corrective measures effectively address them.
OpenAI hopes to extend this approach through partnerships with national AI safety institutes, enabling researchers to test models against real-world security challenges and share findings that improve system reliability.
Protecting Public Services From Emerging AI Risks
NASK’s research also focuses on protecting public services and Poland’s digital information environment. Its AI safety work includes developing safeguards against emerging threats, detecting synthetic AI-generated content, and investigating how malicious actors exploit generated material.
Multilingual analysis is an important part of this work, particularly as AI-generated misinformation and other forms of manipulated content spread across different languages and digital platforms.
NASK Director Radosław Nielek emphasized the importance of cooperation between AI developers and independent research institutions, arguing that AI safety requires coordinated efforts rather than isolated research.
OpenAI’s European policy leadership similarly highlighted the value of external expertise in challenging assumptions, evaluating model behavior, and strengthening safeguards as AI capabilities advance.
Expanding International AI Safety Cooperation
The agreement adds to OpenAI’s existing collaborations with international institutions, including the UK AI Security Institute and the US Center for Advancing Standards and Innovation for Super Intelligence (CAISSI).
According to OpenAI, these partnerships have helped identify and address security vulnerabilities in AI agents. The company has also advocated for greater international coordination on technical standards governing advanced AI systems.
The work with national institutes is expected to cover several areas, including capability evaluations, risk assessments, safeguard testing, and incident reporting. Shared research and testing practices could help participating countries compare findings and coordinate responses to risks that extend across national borders.
Building More Reliable AI Systems
The partnership with NASK reflects a broader effort to make AI safety research more collaborative and operational. Rather than focusing exclusively on theoretical risks, researchers can use advanced models to investigate concrete vulnerabilities, test safeguards, and verify whether security fixes work as intended.
The RouterOS investigation provides an example of how this process can translate research into practical security improvements. As AI systems become more capable, cooperation between developers, public institutions, and cybersecurity specialists will be increasingly important for identifying emerging threats and establishing consistent safety standards.
By expanding access to advanced models and supporting joint technical evaluations, OpenAI and NASK aim to contribute to international efforts to develop AI systems that are more secure, reliable, and resilient.
