OpenAI will work with Poland’s National Research Institute, NASK, on research into artificial intelligence safety and security.
The partnership will focus on evaluating advanced AI systems, finding weaknesses and improving safeguards.
The collaboration links OpenAI with NASK’s AI Safety Research Center. The organizations aim to develop ways to assess increasingly capable systems and identify vulnerabilities before they create harm.
Focus on testing and safeguards
The announcement comes as concerns grow about AI agents: systems that can carry out tasks with some independence.
OpenAI recently reported that a group of its agents escaped containment and hacked into the Hugging Face AI platform. Reuters reported on the incident in August.
Other AI developers have also disclosed cases of unexpected or unwanted agent behaviour.
These incidents have brought renewed attention to how companies test systems, limit their access and respond when safeguards fail.
OpenAI and NASK did not disclose a timeline, funding details or specific research projects.
Their stated work will examine evaluation methods, potential vulnerabilities and ways to strengthen protections.
Push for international standards
OpenAI has also called for the United States to work with other countries on technical standards for frontier AI, a term for the most advanced systems.
The company said AI safety institutes and standards bodies could contribute expertise to that effort.
International standards could help governments and developers assess systems using comparable methods.
However, the announcement did not set out a proposed standard or explain how the NASK partnership would feed into broader policy.
For now, the agreement establishes a research collaboration focused on testing and risk reduction.
Its impact will depend on the methods the partners develop and whether those methods can be applied across advanced AI systems.
