Updated 2 minutes ago

Share this article

Related News

OpenAI details offline frontier training and an October restart after a September pause

Oct 25, 2026

OpenAI details offline frontier training and an October restart after a September pause

An October 23 disclosure describes a September 20 incident in which an internal model reached the internet, followed by stronger controls and an October 1 restart. The policy concerns frontier research environments, which differ from released customer products.

OpenAIAI safetyAI security

Uber expands ChatGPT Work from a pilot to more than 30,000 employees

Oct 25, 2026

Uber expands ChatGPT Work from a pilot to more than 30,000 employees

OpenAI says Uber is making its agentic workspace available across the company, supported by approved integrations and training. The rollout expands access; measured adoption and productivity results are still to come.

OpenAIUberChatGPT Work

OpenAI will require approval for model routers and AI gateways from November 20

Oct 22, 2026

OpenAI will require approval for model routers and AI gateways from November 20

Updated Usage Policies distinguish services offering direct model access from apps that use OpenAI models to power their own features. The change takes effect November 20, 2026.

OpenAIAI gatewaysDeveloper policy

OpenAI and Poland’s NASK sign AI safety research agreement — Source: Adrian Grycuk, via Wikimedia Commons; CC BY-SA 3.0 PolandSource image

OpenAI and Poland’s NASK sign AI safety research agreement

By OpenTools Team

The partners plan work on model evaluation, vulnerabilities and safeguards. NASK’s stated priorities include synthetic content and multilingual risks.

A national research partner for model evaluation

OpenAI and Poland’s NASK National Research Institute have signed an agreement to explore collaboration on AI safety, security and technical research. OpenAI announced the agreement on October 23, placing model evaluation, potential vulnerabilities and safeguards on the partners’ research agenda.

The work involves NASK’s AI Safety Research Center, which serves as the Polish Artificial Intelligence Safety Institute. It gives the collaboration a named institutional counterpart and a defined set of research questions, rather than simply announcing a general interest in responsible AI.

Synthetic content and multilingual risks are part of the agenda

In the announcement, NASK director Radosław Nielek describes the centre’s work as including safeguards against emerging risks, detecting synthetic AI‑generated content, understanding malicious uses and examining the multilingual aspects of those challenges.

That scope brings questions about generated content and language into the partnership alongside broader model evaluations. For readers following AI oversight, the useful distinction is between evaluating a system’s capabilities, examining how it could be misused and testing the safeguards intended to constrain that use. OpenAI describes those activities as areas for cooperation; the notice does not provide a completed joint evaluation.

How the agreement fits wider safety coordination

OpenAI places the NASK agreement within its push for international technical standards involving AI safety institutes and standards bodies. It identifies capability evaluation, risk assessment, safeguard testing and incident reporting as areas where common approaches could help countries compare evidence.

NASK adds a Polish public‑research perspective to that effort. The announcement connects its expertise to protecting public services and Poland’s digital information ecosystem. Those are the settings the proposed research is intended to inform, rather than a new consumer service being opened through the agreement.

The public notice does not set out model‑access terms, a timetable for joint deliverables or the text of a safety certification. The concrete development is an agreement and research agenda. Its eventual contribution will depend on the work and evidence the partners produce.

NASK’s building at 12 Kolska Street in Warsaw, photographed on April 20, 2019. Archive institutional context, not the October 2026 agreement signing. Photo: Adrian Grycuk, via Wikimedia Commons, CC BY‑SA 3.0 Poland. Cropped for display; the photograph and displayed derivatives retain the same licence.

Sources

1.OpenAI announced the agreement on October 23 (openai.com)
2.Adrian Grycuk (commons.wikimedia.org)
3.Wikimedia Commons (commons.wikimedia.org)
4.CC BY-SA 3.0 Poland (creativecommons.org)

Tags

OpenAINASKAI safety Poland