OpenAI and Poland’s NASK sign an agreement to study how capable systems fail
OpenAI’s post, dated Friday: the institute’s AI Safety Research Center, which acts as Poland’s safety institute, and OpenAI will look at evaluations, vulnerabilities and safeguards. No budget or timetable is in the post. It points back to July, when CERT Polska used Daybreak access to help find six RouterOS bugs the vendor later fixed.
Short1 minEasy
Eli Navarro1 min read
- OpenAI
- NASK
- Poland
- Safety
- CERT Polska

Photo: OpenAISource page
OpenAI said on Friday it has signed an agreement with Poland’s NASK National Research Institute to explore work on AI safety, security and technical research. The post, dated 23 October, says NASK’s AI Safety Research Center — the body that serves as Poland’s artificial intelligence safety institute — and OpenAI will look at ways to evaluate more capable systems, find vulnerabilities and strengthen safeguards. Reuters reported the agreement on Friday. The post names no budget, no timetable, no shared cluster and no model that NASK would receive under the new agreement.
Radosław Nielek, the director of NASK, called the collaboration a natural step for an institute already working on safeguards, on spotting synthetic content, and on malicious uses and the multilingual side of those problems. “AI safety cannot be achieved in isolation,” he said. Tom Duff Gordon, OpenAI’s vice president and head of EMEA policy, said many of the people who have helped lead OpenAI’s research are in Poland, and that the agreement “opens the door” to using that expertise to test the company’s systems. The post then points to July, not to a project that starts today. CERT Polska, which sits inside NASK, received access then to OpenAI’s cyber-capable models through a program called Daybreak. OpenAI says the team used the models to help 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. That paragraph is OpenAI’s account of work already done. It does not describe a test plan under Friday’s signature.


