OpenAI announces a research collaboration with Poland’s NASK National Research Institute focused on AI safety and security. The agreement is aimed at studying how increasingly capable AI systems can behave in unintended ways, and at building stronger safeguards.

Across the two reports provided, both outlets describe the work as involving testing more powerful AI models, looking for vulnerabilities, and developing mitigation measures. The Next Web frames the partnership around growing concerns about AI agents not behaving as developers expect. Channel NewsAsia presents the same basic development—OpenAI teaming with NASK on AI safety research—without additional specifics in the excerpt provided.

Overall, the coverage aligns on the partnership’s purpose and the general areas of work (testing, vulnerability identification, and safeguards), while the emphasis differs slightly: one outlet foregrounds the risk of unintended agent behaviour, while the other sticks more closely to the collaboration itself.