OpenAI and Poland NASK Partner on AI Safety Research

Introduction

OpenAI and Poland's NASK National Research Institute have announced a strategic partnership focused on advancing AI safety, security, and technical research. The collaboration brings together two organizations with complementary expertise to address emerging risks in artificial intelligence and build stronger safeguards for increasingly capable systems.

As AI systems grow more sophisticated, the need for coordinated international efforts to understand risks and develop robust safety measures becomes critical. This partnership reflects a growing trend of cooperation between leading AI developers and national research institutions.

Research Collaboration and Objectives

The agreement establishes a framework for joint research between NASK's AI Safety Research Center, which serves as the Polish Artificial Intelligence Safety Institute, and OpenAI's research teams. The collaboration will focus on evaluating advanced AI systems, identifying potential vulnerabilities, and strengthening safety protocols.

Key areas of joint investigation include methods for assessing AI capabilities, developing risk evaluation frameworks, and creating standardized approaches to testing safeguards. The partnership also aims to advance incident reporting mechanisms and establish common methodologies for comparing evidence across different AI systems and applications.

Building on Existing Cybersecurity Work

This new research collaboration builds upon previous security-focused work between OpenAI and Polish cyber defense teams. In July, CERT Polska, which operates within NASK, gained access to OpenAI's cyber-capable models through the Daybreak program.

Using these models, CERT Polska researchers successfully identified six vulnerabilities in MikroTik RouterOS, including flaws that could be combined to achieve unauthenticated remote control of affected devices. The vendor subsequently released patches, which CERT Polska verified effectively prevent the attack scenarios they had discovered.

International Standards and Technical Coordination

OpenAI has been advocating for international cooperation on technical standards for frontier AI development, emphasizing the importance of involving AI safety institutes and standards bodies from multiple countries.

The organization argues that common approaches to capability evaluation, risk assessment, and safeguard testing enable nations to share evidence and coordinate responses to threats that transcend borders. The partnership with NASK demonstrates how national expertise can be integrated into broader global safety initiatives.

NASK's extensive research in AI safety, particularly as it applies to protecting public services and Poland's digital information infrastructure, provides valuable real-world context that can inform global standards development.

What Security Teams Should Do Now

Organizations working with or developing AI systems should take several immediate steps to prepare for evolving safety standards:

• Track emerging standards: Monitor developments from collaborations like the OpenAI-NASK partnership, as their research methodologies may become adopted industry practices.
• Strengthen incident response: Develop capabilities for detecting, reporting, and responding to AI-related security incidents, including those involving model vulnerabilities.
• Evaluate existing systems: Begin assessing current AI deployments against emerging safety frameworks, particularly around capability evaluation and risk assessment.
• Engage with national initiatives: Connect with local AI safety institutes and cybersecurity agencies to stay informed about regional developments and requirements.

Conclusion

The OpenAI-NASK partnership signals a growing recognition that AI safety requires coordinated international effort between developers and public research institutions. As these collaborations mature, they will likely produce the technical standards and evaluation methods that shape the future of AI development and deployment. Organizations seeking to maintain secure and compliant AI operations should consider how their own security practices align with these emerging frameworks, and Grivyonx's AI automation and cloud security services can help bridge the gap between evolving standards and practical implementation.