What happened
OpenAI and Poland's NASK National Research Institute signed an exploratory agreement to collaborate on and cybersecurity research. The partnership involves NASK's AI Safety Research Center, which functions as Poland's national AI Safety Institute. The agreement focuses on evaluating capable AI systems, identifying security vulnerabilities, and strengthening safeguards. This follows earlier concrete cooperation where CERT Polska used OpenAI models to identify six vulnerabilities in MikroTik RouterOS, which were subsequently fixed. OpenAI confirmed the deal via LinkedIn, noting it extends its work with other public AI safety institutes like the UK's AISI and the US's CAISSI.
OpenAI and Poland's NASK National Research Institute have signed an agreement to explore joint research on artificial intelligence safety and cybersecurity. The collaboration involves NASK's Research Center, which serves as the Polish Artificial Intelligence Safety Institute (AISI). The primary focus of the agreement is to examine methods for evaluating increasingly capable AI systems, identifying security vulnerabilities, and strengthening safeguards against emerging threats.
This partnership builds on existing cybersecurity work between the two organizations. Previously, CERT Polska, the computer emergency response team operating within NASK, used OpenAI models to uncover six vulnerabilities in MikroTik RouterOS software. According to OpenAI, some of these vulnerabilities could be combined to allow attackers to take control of affected devices without authentication. MikroTik subsequently released fixes for these flaws. The new agreement broadens this cooperation from using AI for cybersecurity to evaluating the safety and security of the AI systems themselves.
George Osborne, Head of OpenAI for Countries, confirmed the agreement on LinkedIn, highlighting NASK's experience in protecting public infrastructure. He stated that NASK brings experience protecting public services, helping ground international safety research in the challenges institutions face in practice. Tom Duff Gordon, VP and Head of EMEA Policy at OpenAI, added that as AI becomes more capable, it is vital to draw on expertise beyond the organization to challenge assumptions and test systems.
The agreement is explicitly described as exploratory, meaning the organizations are identifying areas for potential technical collaboration rather than announcing the completion of a specific joint research project. NASK brings expertise in cybersecurity research, detecting AI-generated synthetic content, investigating malicious AI use, and addressing multilingual security challenges. The institute also supports Poland's public sector through digital infrastructure projects and cybersecurity training.
Source details: edtechinnovationhub.com
Why it matters
This agreement signals a growing trend of frontier AI developers formalizing partnerships with national public research institutions to address safety and security risks. By involving a national institute, OpenAI gains access to localized expertise in multilingual security and public infrastructure protection. It also provides NASK with direct access to frontier model capabilities for safety evaluation, potentially enhancing Poland's capacity to assess AI risks within its own digital environment. This move supports the broader industry push toward international technical standards for and incident reporting.
The agreement represents a concrete step in the integration of frontier AI developers with national public research institutions. By partnering with NASK, OpenAI extends its existing work with other public organizations, including the UK's AI Security Institute and the US Center for Advancing Standards and Innovation for Super Intelligence (CAISSI). This suggests a strategic move to align with international technical standards for frontier AI, including shared approaches to evaluating model capabilities and reporting security incidents.
For Poland, the partnership provides access to advanced AI capabilities for safety research, potentially enhancing the nation's ability to assess AI risks in its specific digital and linguistic context. NASK's focus on multilingual security challenges is particularly relevant, as risks may differ depending on the languages and digital environments in which AI systems operate. This collaboration could lead to more robust safety evaluations that account for local nuances.
The prior success in identifying vulnerabilities in MikroTik RouterOS demonstrates the practical utility of this type of collaboration. It shows that AI models can be effectively used to uncover security flaws in critical infrastructure software. The new agreement aims to apply similar rigorous testing to the AI systems themselves, which is a critical area of concern as models become more capable and widely deployed.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Why can ethical evaluation not be reduced to one model score?
What to watch next
Watch for specific joint research outputs or publications resulting from this exploratory framework. Monitor whether this leads to standardized safety evaluation protocols adopted by other national institutes. Also observe if similar agreements are signed with other national bodies, indicating a broader shift in how AI safety research is conducted globally.
Monitor for the publication of specific research findings or technical reports resulting from the OpenAI-NASK collaboration. Since the agreement is exploratory, the first tangible outputs will likely be joint papers or case studies on evaluation methods.
Observe whether this partnership leads to the adoption of standardized safety evaluation protocols by other national institutes. If OpenAI and NASK develop a framework that is widely adopted, it could influence global AI safety standards.
Watch for similar agreements between other frontier AI developers and national research institutions. This could indicate a broader trend toward public-private partnerships in , potentially shifting the landscape of how AI risks are managed and regulated internationally.