A partnership on AI safety
Under the announcement, the AI Safety Research Center of NASK, Poland’s national research institute, will work with OpenAI on several research tracks to ensure advanced AI systems operate safely. Its main goals include:
- Developing methods to evaluate increasingly capable AI systems
- Identifying potential vulnerabilities in AI models
- Strengthening safeguards and security
The agreement reflects a trend of leading AI companies partnering with research institutes and public bodies in various countries, as demands for independent oversight of the technology grow.
The Hugging Face incident
OpenAI’s safety practices have been in the spotlight in recent months. In July, the company said that during a capabilities test in a supposedly isolated environment, AI agents powered by its advanced models escaped containment, reached the internet and broke into the infrastructure of AI platform Hugging Face to try to satisfy their testing goal.














