OpenAI and Poland's NASK sign AI safety agreement: evaluating stronger systems, hunting vulnerabilities, and plugging into Polish public cyber defense
Synopsis
OpenAI and Poland's National Research Institute NASK have signed an agreement to collaborate on AI safety, security and technical research. NASK's AI Safety Research Center, which serves as the Polish Artificial Intelligence Safety Institute, and OpenAI will explore ways to evaluate increasingly capable AI systems, identify potential vulnerabilities and strengthen safeguards. The agreement extends OpenAI's work with public research institutions including the US CAISSI and the UK AISI, and builds on NASK's CERT Polska gaining access to OpenAI cyber-capable models through Daybreak and using them to find six vulnerabilities in MikroTik RouterOS.
Interpretation
OpenAI and Poland's National Research Institute NASK have signed an agreement covering collaboration on AI safety, security and technical research, aimed at understanding emerging risks, strengthening safeguards and exploring how advanced AI can support a more secure and resilient digital ecosystem. Earlier collaborations included public research institutions such as the US Center for Advancing Standards and Innovation for Super Intelligence (CAISSI) and the UK AI Security Institute (AISI), work that has been used to strengthen safeguards, identify and fix security vulnerabilities in AI agents, and advance frontier AI safety. The agreement specifies that NASK's AI Safety Research Center, serving as the Polish Artificial Intelligence Safety Institute, and OpenAI will collaborate on AI safety research, exploring ways to evaluate increasingly capable AI systems, identify potential vulnerabilities and strengthen safeguards.
The technical focus is evaluating increasingly capable AI systems, finding potential vulnerabilities and hardening safeguards, while NASK also works on detecting synthetic AI-generated content and understanding its malicious uses and multilingual aspects. NASK's research is applied directly to protecting public services and Poland's digital information ecosystem, bringing national practice into the wider effort on international AI safety standards. In a statement on the agreement, NASK Director Radosław Nielek lists the center's research scope: safeguards against emerging risks, detection of synthetic AI-generated content, and understanding its malicious uses and multilingual challenges.
The agreement follows OpenAI's existing work with cyber defenders in Poland: in July 2026, CERT Polska, which operates within NASK, gained access to OpenAI's cyber-capable models through Daybreak. The team used OpenAI models to discover six vulnerabilities in MikroTik RouterOS, including flaws that could be combined to take control of affected devices without authentication. The vendor released fixes, which CERT Polska confirmed prevented the attacks it observed.
Perspective
The agreement is at the stage of exploring collaboration; the public content is the direction of cooperation and statements from both sides, with no specific research projects, timelines or evaluation results given. The concrete technical record shown is CERT Polska using OpenAI models through Daybreak to find six vulnerabilities in MikroTik RouterOS, with the vendor releasing fixes.
The agreement describes collaboration to be carried out; evaluation methods, vulnerability-identification procedures and concrete safeguard outputs have not been published. OpenAI's statement that existing collaborations have already helped strengthen safeguards is an institutional claim without quantitative results.