SITUATION REPORT: Opportunity, moderate significance, long tail.
OpenAI is partnering with NASK, Poland’s national research institute for cybersecurity and digital infrastructure. According to OpenAI, the two will work together on AI safety, security, and technical research. The stated goal is to “strengthen safeguards and support a more resilient digital ecosystem.”
The announcement is short. OpenAI hasn’t shared timelines, budgets, or named research projects. Still, the choice of partner tells you plenty, and it fits a pattern worth tracking.
🎯 Who NASK Is
NASK isn’t a startup or a think tank. It sits at the center of Poland’s digital defenses. If you’ve never heard of it, these are the points that count:
- It runs CERT Polska. That’s one of Poland’s main national teams for responding to cyber incidents. They deal with phishing, malware, and attacks on critical systems every day.
- It manages the .pl domain registry. So it has a direct view of how fraud and abuse spread across the Polish internet.
- It builds language models. NASK helped develop PLLuM, a Polish-language large language model backed by the state. So it knows AI from the inside, not just as a threat to defend against.
- It’s on the front line. Poland borders Ukraine and Belarus, and it’s become one of Europe’s most targeted countries for cyberattacks and disinformation.
That mix is rare. Most AI safety partners are either strong in research or strong in operations. NASK does both.
📡 Why This Matters
AI labs have spent the past two years building ties with governments on safety testing. Most of that work went through dedicated AI safety institutes in the US and UK. This deal is different. It connects OpenAI to an organization that handles real security incidents as they happen.
What stands out here is the word “security” sitting next to “safety.” Safety research asks whether a model behaves as intended. Security research asks how attackers will abuse it. Think AI-written phishing in Polish, deepfake scams aimed at local banks, or influence campaigns ahead of elections. A national CERT sees those attacks first. That gives OpenAI real-world signal it can’t get from lab red-teaming alone.
The benefit runs both ways. NASK gets closer access to a frontier lab’s technical know-how. Poland gets a stronger voice in how these systems handle its language and its threats.
🗺️ Context: The Bigger Picture
This fits a trend that’s been building fast:
- Labs are going local. OpenAI and its rivals keep signing country-level deals for education, government services, and infrastructure. Safety partnerships are the next layer.
- Europe wants a say. Under the EU AI Act, providers of general-purpose AI models carry risk and security obligations. Partnerships with trusted national bodies help show those obligations being met.
- Smaller languages need attention. Safeguards built mostly on English data can miss abuse in Polish, Czech, or Lithuanian. Local partners help close those blind spots.
Central and Eastern Europe has become a proving ground. It’s highly digital, under heavy attack, and very motivated to defend itself.
⚙️ What Practitioners Should Watch
- Published research. If this partnership produces joint papers or threat reports, expect useful data on how attackers actually use AI in non-English settings.
- Better abuse detection. Teams building on OpenAI’s models in Poland, and maybe the wider region, could see stronger protection against local fraud patterns.
- Copycat deals. Other national CERTs and cyber agencies will likely seek similar arrangements. Other AI labs will probably compete for them too.
- Policy echoes. Findings from a group like NASK can shape how EU regulators think about AI security requirements.
If you’re a security lead at a European company, this is a cue to check how your incident response handles AI-enabled attacks. The organizations defending national networks are clearly treating it as a priority.
🔭 Outlook
For now this is a statement of intent, not a finished program. The real test is output: shared threat intelligence, published findings, and measurable improvements in safeguards. If the partnership delivers, it could become a template for how frontier labs work with national security bodies across Europe.
Full details of the announcement are available from OpenAI.