OpenAI NASK partnership marks a formal collaboration between the AI developer and Poland’s NASK National Research Institute to explore AI safety, security and technical research. The agreement brings together NASK’s national research capabilities and OpenAI’s experience with public research institutions and standards bodies to better understand emerging risks, strengthen safeguards and study how advanced AI can support a more resilient digital ecosystem.
Under the terms of the agreement, NASK’s AI Safety Research Center—which serves as the Polish Artificial Intelligence Safety Institute (AISI)—and OpenAI will cooperate on research that evaluates increasingly capable AI systems, identifies potential vulnerabilities and seeks to improve safeguards for deployment and use. The partners say their work will focus on practical, technical research to assess model capabilities and to test and harden defenses against risks that can arise as systems grow more sophisticated.
Planned collaboration areas include capability evaluation, methods to surface weaknesses in models, testing of safeguards in real-world contexts and mechanisms for incident reporting. The agreement frames these activities as part of a broader effort to establish common approaches for evaluating AI systems, assessing risk and ensuring that protections are effective in practice. By combining NASK’s operational perspective—derived from protecting public services and Poland’s digital information ecosystem—with OpenAI’s technical collaborations, the partners aim to ground international safety work in the environments where AI is actually used.
The agreement also situates the collaboration within a wider advocacy by OpenAI for international coordination on technical standards for frontier AI. OpenAI has previously called for cooperation between the US and other countries that draws on the expertise of AI safety institutes and standards bodies; this partnership with a national research institute is presented as an extension of that approach. OpenAI already works with groups such as the US Center for Advancing Standards and Innovation for Super Intelligence (CAISSI) and the UK AI Security Institute (also abbreviated AISI in the UK), and the new agreement with NASK expands its network of public research collaborations.
A concrete line of prior engagement between OpenAI and Polish cyber defenders underpins the partnership. In July, CERT Polska, which operates within NASK, obtained access to OpenAI’s cyber-capable models through the Daybreak program. Using those models, CERT Polska identified six vulnerabilities in MikroTik RouterOS, including flaws that could be chained to gain control of affected devices without authentication. The vendor issued fixes for the reported issues, and CERT Polska confirmed the updates prevented the observed attacks. That work is cited by both organizations as an example of how AI-assisted research and defensive collaboration can lead to real-world mitigations.
The agreement does not enumerate specific timelines or deliverables. Instead, it signals an ongoing technical exchange between a national research institute and a major AI developer focused on the security and safety implications of increasingly capable systems. The partners describe the collaboration as a mechanism for ensuring national expertise informs global efforts to manage AI risk, including creating common technical standards and shared approaches for capability evaluation, testing safeguards and incident reporting.
For Poland, integrating NASK’s applied safety research into international technical work is intended to keep global standards connected to national operational realities—how public services and the country’s digital infrastructure are protected. For OpenAI, the partnership extends its engagement with public research institutions and standards bodies, positioning national institutes as contributors to internationally coordinated safety work.
Both organizations emphasize that international cooperation between developers and public research institutions is essential for understanding and mitigating cross-border risks as AI systems become more capable. Shared technical standards and evidence-based approaches to model evaluation and incident reporting can help countries compare results and respond to transnational threats more effectively.
The OpenAI NASK partnership therefore represents a deliberate effort to bridge national operational expertise with international technical work on AI safety and security. While many details remain to be defined, the agreement establishes a framework for collaborative research, testing and defensive practice that seeks to keep pace with the technical challenges posed by advancing AI systems.
As the partnership proceeds, its measurable outcomes—such as joint research reports, shared testing methodologies or coordinated incident responses—will determine how effectively this model of collaboration can inform broader international standards and practical safeguards for AI deployment.
Source: Read the original source
Leave a Reply