OpenAI and Poland’s NASK join forces on AI safety research

National expertise meets frontier-model testing

OpenAI and Poland’s NASK National Research Institute have agreed to explore collaboration on AI safety, security and technical research. NASK’s AI Safety Research Center serves as Poland’s AI Safety Institute, giving the partnership a public-interest counterpart with local scientific, cyber and information-integrity expertise. The organisations plan to study emerging risks, evaluate increasingly capable systems, identify vulnerabilities and strengthen safeguards.

The agreement reflects a wider shift from safety commitments made by a developer alone towards testing with national research institutions. External expertise can challenge assumptions and bring multilingual, cultural and operational contexts that a global model provider may not reproduce internally. The value will depend on whether researchers receive meaningful access, can publish useful findings and have a route for unresolved concerns to influence deployment decisions.

Cyber work provides an early practical example

OpenAI points to existing collaboration with cyber defenders in Poland. CERT Polska, which operates within NASK, gained access to cyber-capable OpenAI models through Daybreak in July. The team used the models while investigating MikroTik RouterOS and found six vulnerabilities, including flaws that could be combined to take control of affected devices without authentication. The vendor issued fixes and CERT Polska confirmed that the observed attacks were prevented.

That example illustrates both benefit and risk. Advanced models can expand a defender’s capacity to find weaknesses, yet the same capabilities could reduce the effort required by attackers. Access therefore needs verified users, scoped environments, activity monitoring and coordinated disclosure procedures. Success should be judged by responsibly resolved vulnerabilities and improved defensive readiness, not simply by the number of findings a model produces.

Shared methods need measurable outcomes

OpenAI says international cooperation can support common approaches to capability evaluation, risk assessment, safeguard testing and incident reporting. Comparable methods would make it easier for governments and developers to share evidence about risks that cross borders. They should also preserve differences in law, language and public-service conditions rather than forcing every country into a single developer-defined benchmark.

A credible program needs pre-agreed research questions, reproducible evaluation sets and clear disclosure rules. Both organisations should distinguish exploratory research from independent assurance and report the limitations of their work. The partnership is a promising route to broader scrutiny, but public trust will grow only if collaboration produces evidence that outsiders can assess and safeguards that demonstrably change model or product behaviour.

What teams should validate before production

For implementation, teams should record the exact model, software, dataset, connector and policy versions behind every result. Representative evaluation must include normal work, boundary cases, deliberate misuse and component failure, with outcomes separated by user role, data sensitivity and consequence. A strong average should never hide a serious failure in a critical path. Changes should pass the same regression set before promotion, and operators need a documented fallback that preserves evidence rather than silently broadening access or lowering quality.

Procurement and planning should account for the human work required to review exceptions, maintain controls and investigate incidents. Published performance or efficiency claims are useful starting points, but production confidence comes from evidence collected in the organisation’s own environment. A named owner, measurable acceptance threshold, cost model and tested shutdown path turn a promising release into a service that can be governed over time. Where decisions affect safety, rights, money or regulated records, accountable professional review remains essential.

Security and operational evidence

Security review should map every identity, credential, network destination, stored artefact and administrative access path involved in the workflow. Logs need enough context to reconstruct who initiated an action, which evidence was consulted, what the system proposed and what ultimately executed. Sensitive prompts and traces should not be retained merely because they are useful for debugging. Retention, redaction and access must follow the underlying data classification, with a practical process for correcting or removing material when policy or law requires it.

Operational monitoring should measure quality and service behaviour together. Latency, throughput and availability matter, but so do unsupported claims, missed constraints, unsafe tool calls and the rate at which people overturn recommendations. Teams should rehearse dependency outages and partial failures before unattended operation. If a model, connector or external service becomes unavailable, the workflow should fail predictably, preserve its state and tell the user what remains unfinished. Clear degradation is safer than an agent improvising around a missing control.

Transparency should scale with adoption

Organisations should explain the scope and limitations of the deployed capability to the people affected by it. Users need to know when automation is involved, how to challenge an outcome and where a human can intervene. That transparency improves feedback and prevents a pilot result from being interpreted as blanket approval for broader autonomy. Expansion should follow demonstrated evidence, not novelty or pressure to match a competitor announcement. The durable advantage comes from dependable integration, careful measurement and learning that survives the next model or product update.

Finally, teams should schedule a review after real usage rather than treating launch as the end of governance. The review should compare expected benefits with observed quality, cost, incidents and user behaviour, then retire workflows that do not justify their complexity. A service that works well can expand through the same evidence-based process. A service that fails should be narrowed or stopped without losing the records needed to understand why. This feedback loop is what turns experimentation into accountable operational capability.