OpenAI has signed an agreement with Poland’s National Research Institute (NASK) to explore collaboration on artificial intelligence safety, cybersecurity, and technical research. The partnership will give Polish public researchers access to advanced AI models and testing workflows to help evaluate emerging risks and strengthen safeguards.
The initiative will bring together OpenAI and NASK’s AI Safety Research Center, which also serves as Poland’s Artificial Intelligence Safety Institute (AISI). Their work will focus on evaluating increasingly capable AI systems, identifying potential vulnerabilities, and improving protections against emerging threats.
Cybersecurity Testing Identifies Router Vulnerabilities
One of the partnership’s most significant examples of collaboration took place before the agreement was signed.
In July, CERT Polska, the incident response team operating within NASK, gained access to OpenAI’s cyber-capable models through a channel called Daybreak. Researchers used the models to examine MikroTik RouterOS and identified six vulnerabilities.
The security flaws could be combined to take control of affected devices without requiring authentication, potentially exposing vulnerable network infrastructure to unauthorised access.
MikroTik subsequently released security updates addressing the vulnerabilities. CERT Polska then tested the patches and confirmed that they prevented the attack methods researchers had identified against unpatched devices.
The process demonstrated how advanced AI models can support cybersecurity research by helping analysts identify vulnerabilities, validate attack techniques, and provide vendors with actionable information for remediation.
OpenAI hopes to replicate this type of testing and collaboration with national AI safety institutes in other countries.
Strengthening AI Safeguards for Public Services
NASK’s research focuses on applying AI safety techniques to the protection of public services and Poland’s wider digital information environment.
The institute’s responsibilities include identifying synthetic AI-generated content and examining how such material can be used maliciously. This includes research into multilingual threats, where AI-generated content can spread across different languages and create additional challenges for detection and monitoring.
NASK Director Radosław Nielek described the agreement as a natural extension of the institute’s existing AI safety research, highlighting the importance of cooperation between public research institutions and developers of advanced AI systems.
He emphasised that AI safety requires collaboration across organisations, particularly as emerging technologies introduce risks that cannot be addressed by individual developers or governments working independently.
Tom Duff Gordon, OpenAI’s vice president and head of EMEA Policy, similarly highlighted the importance of independent expertise in evaluating increasingly capable AI systems.
The partnership is intended to allow researchers to challenge existing assumptions, test AI systems under different conditions, and identify opportunities to strengthen safeguards.
Expanding International AI Safety Collaboration
The agreement builds on OpenAI’s existing relationships with other national AI safety organisations, including the UK’s AI Security Institute and the US Center for Advancing Standards and Innovation for Super Intelligence (CAISSI).
OpenAI says its previous collaborations with these institutions have helped identify and address security vulnerabilities in AI agents.
The company has also advocated for greater international coordination on technical standards for advanced AI systems, arguing that governments, research institutes, and technology developers need shared approaches to evaluating emerging risks.
OpenAI has outlined several areas for cooperation with national institutes, including capability evaluations, risk assessments, safeguard testing, and incident reporting.
These activities are intended to establish more consistent methods for measuring AI capabilities and identifying vulnerabilities. Shared reporting practices could also help researchers and governments compare findings when security risks extend across national borders.
Building More Reliable and Secure AI Systems
The partnership between OpenAI and NASK reflects the growing importance of independent AI safety research as advanced models become more capable and widely deployed.
The CERT Polska RouterOS investigation provides a practical example of how access to advanced AI models can support vulnerability discovery and security testing, with identified issues progressing through vendor remediation and validation.
Moving forward, cooperation between AI developers and national research institutions could help establish stronger testing practices, improve transparency around emerging risks, and strengthen safeguards across different applications.
As AI systems become increasingly integrated into critical infrastructure and public services, international research partnerships will play an important role in ensuring that technological progress is accompanied by effective security measures and responsible oversight.
