OpenAI has signed an agreement with Poland’s NASK National Research Institute to explore new ways of improving AI safety, strengthening cybersecurity and evaluating the risks posed by increasingly capable artificial intelligence systems.
The partnership brings together OpenAI’s technical expertise and NASK’s research capabilities to help identify vulnerabilities, assess emerging threats and develop stronger safeguards for advanced AI models. The two organizations will also explore how their combined knowledge can contribute to a safer and more resilient digital environment.
As AI systems become more capable, ensuring that they remain reliable and secure is becoming an increasingly important challenge. Collaboration between technology developers, public research institutions and national cybersecurity organizations could help identify risks earlier and establish more consistent methods for testing AI systems.
Through the agreement, OpenAI will work with NASK’s AI Safety Research Center, which serves as Poland’s Artificial Intelligence Safety Institute. Their collaboration will focus on AI safety research, evaluating advanced systems, identifying potential weaknesses and exploring methods to improve protective measures.
OpenAI and NASK join forces on AI safety research
The agreement reflects a broader effort to bring AI developers and independent research institutions closer together as the technology advances.
OpenAI develops increasingly capable AI systems, while NASK conducts research and technical work related to cybersecurity, digital resilience and the protection of Poland’s information ecosystem.
By combining their expertise, the organizations aim to investigate how advanced AI systems behave, identify potential vulnerabilities and improve the safeguards designed to reduce risks.
The collaboration is also expected to examine ways to evaluate AI capabilities and determine whether existing protections remain effective as models become more sophisticated.
These evaluations are important because traditional testing methods may not always capture the full range of risks associated with advanced AI. Researchers need ways to assess how systems respond to difficult situations, identify weaknesses and understand how malicious actors might attempt to misuse them.
NASK’s AI Safety Research Center will contribute its research experience, including work on emerging AI threats and the detection of synthetic content. This includes understanding how AI generated material can be misused and how challenges differ across languages.
Such research is particularly relevant in multilingual environments, where harmful or misleading AI generated content can be harder to identify consistently.
The agreement establishes a framework for cooperation, although the announcement does not specify a detailed timetable, funding arrangement or a complete list of research projects that will follow.
Poland’s research expertise will support international AI security efforts
Poland’s involvement highlights the importance of including national research institutions in the development of AI safety practices.
AI risks do not stop at national borders. A vulnerability in a widely used system, a cybersecurity threat or the spread of synthetic misinformation can affect organizations and communities across multiple countries.
For that reason, developing effective safeguards requires more than the efforts of individual technology companies. Public research institutions can contribute independent expertise, local knowledge and practical experience of the threats facing their countries.
NASK’s work is closely connected to protecting public services and Poland’s digital information environment. This gives the institute an opportunity to examine AI risks in settings where security, reliability and public trust are essential.
The partnership could help connect international AI safety research with the practical challenges faced by national cybersecurity teams.
OpenAI has previously supported collaboration with public research institutions, including the UK’s AI Security Institute and US organizations working on AI standards and security. Its work with these bodies has included efforts to strengthen safeguards, investigate vulnerabilities in AI agents and advance the evaluation of advanced systems.
The agreement with NASK extends this approach by bringing additional Polish research expertise into the international discussion.
However, the effectiveness of the partnership will ultimately depend on the research conducted, the findings shared and the extent to which those findings lead to measurable improvements in AI security.
Cybersecurity work has already delivered practical results
The new agreement builds on existing cooperation between OpenAI and cybersecurity specialists in Poland.
In July, CERT Polska, the cybersecurity team operating within NASK, gained access to OpenAI’s cyber capable models through Daybreak.
According to the announcement, the team used these models to help discover six vulnerabilities in MikroTik RouterOS, software used in networking equipment.
The identified flaws included weaknesses that could be combined to take control of affected devices without authentication. Such vulnerabilities can create serious security risks because an attacker may be able to compromise a device without first obtaining valid login credentials.
The vendor subsequently released fixes, and CERT Polska confirmed that the updates prevented the attacks it had observed.
This example illustrates how AI models can support defensive cybersecurity work when used by technical specialists. They can help researchers investigate software, identify possible weaknesses and accelerate parts of the vulnerability discovery process.
However, AI assisted security research still requires human expertise, careful validation and responsible disclosure. Finding a potential vulnerability is only one stage of the process. Researchers must also determine whether it is exploitable, communicate the findings appropriately and ensure that affected systems can be patched.
The CERT Polska work provides a practical example of how collaboration between AI developers and national cybersecurity teams can contribute to improving digital security.
It also offers useful context for the new partnership, which is intended to explore broader questions about AI safety and security.
Shared standards could help make advanced AI safer
OpenAI has also called for the United States to work with international partners on technical standards for advanced AI systems.
Common approaches to evaluating capabilities, assessing risks, testing safeguards and reporting security incidents could help researchers and governments compare findings more consistently.
Without shared methods, different institutions may assess the same AI system using different criteria, making it harder to compare results or identify gaps in protection.
International cooperation could help establish more consistent expectations for testing and reporting, while allowing researchers to exchange evidence about emerging threats.
The collaboration with NASK reflects the value of involving national AI safety institutions in this work. These organizations can help ensure that international standards take account of the practical conditions in which AI systems are deployed.
For example, AI generated misinformation may present different challenges depending on language and local context. Cybersecurity risks can also vary according to the systems and infrastructure used by different organizations.
Bringing these perspectives into research could help make safety evaluations more representative of real world conditions.
At the same time, agreeing on standards is only part of the challenge. Those standards must be applied consistently, updated as AI capabilities change and supported by credible testing.
The announcement does not establish that OpenAI and NASK have already developed a new international standard. Instead, the agreement creates an opportunity to explore research and cooperation that could contribute to the wider effort.
What the OpenAI and NASK partnership means for AI safety
The agreement represents another step towards closer cooperation between AI developers and public research institutions.
For OpenAI, working with NASK provides access to expertise from researchers and cybersecurity specialists who understand the challenges facing Poland’s digital infrastructure. For NASK, collaboration offers an opportunity to engage with the technical capabilities of a leading AI developer and contribute to research on advanced AI systems.
The partnership also highlights a broader principle: AI safety cannot depend on a single company or institution working alone.
Developers can test their own systems, but outside researchers and national security organizations can bring different perspectives, identify overlooked weaknesses and help establish more consistent evaluation methods.
The existing work with CERT Polska demonstrates how this cooperation can produce tangible cybersecurity benefits. The new agreement could extend that approach into a wider range of AI safety questions.
Much will depend on the results of future research and whether the organizations can turn their shared expertise into stronger safeguards, clearer evaluation methods and practical improvements.
For now, the agreement signals a commitment to continued collaboration rather than a completed solution to the challenges of advanced AI. As the technology develops, partnerships between AI companies and public research institutions are likely to remain an important part of efforts to improve security, reliability and public trust.
Follow TechBSB For More Updates
