OpenAI signs an AI-safety research agreement with Poland's NASK institute

Victoria Shinder26 October 2026, 14:481 min read

OpenAI and Poland's NASK National Research Institute signed an agreement on 23 October to collaborate on AI safety, security and technical research. NASK's AI Safety Research Center, which serves as the Polish AI Safety Institute, will work with OpenAI on evaluating increasingly capable systems, identifying vulnerabilities and strengthening safeguards. The deal joins OpenAI's existing arrangements with the US standards body CAISSI and the UK AI Security Institute.

The agreement is not starting from nothing. In July, CERT Polska, which operates inside NASK, gained access to OpenAI's cyber-capable models through the company's Daybreak programme, and used them to help find six vulnerabilities in MikroTik RouterOS, including flaws that could be chained to take over a device. The new research track formalises that kind of work.

What it means

The interesting part is who is doing the testing. A frontier lab grading its own safety work is a closed loop; a national research institute that also runs a country's CERT brings adversarial pressure the lab cannot generate internally, plus a real deployment surface to measure against. NASK's own director framed it as detecting synthetic content and multilingual abuse — problems that look very different outside English. For a developer, the signal is narrower but concrete: the people finding exploitable bugs in shipping network gear are now pointed at the models too, and that is a better test than any internal red team.