UncertAInty – a WIPANO project for reliable AI systems

February 11, 2026

UncertAInty (Uncertainty Quantification in AI Systems for International Compliance) is a WIPANO project funded by the Federal Ministry for Economic Affairs and Climate Action (BMWK) that advances the systematic handling of uncertainty in AI systems. Over a 30-month period, partners from industry, research, and standardization are working together to clearly describe uncertainties in AI, make them measurable, and standardize them internationally as part of the development of ISO/IEC TS 25223.

Lösungskonzept im Projekt DeepQuali zur Verbesserung der Softwarequalität. — © Fraunhofer IESE

Artificial intelligence opens up new possibilities for vehicles, machines, and systems—especially in safety-critical applications. At the same time, due to the nature of data-driven, learning systems, individual results cannot be guaranteed. To ensure safe and legally compliant use, it is therefore crucial to make performance limits transparent and to quantify uncertainty in a traceable manner. This is exactly where UncertAInty comes in.

Last modified: February 11, 2026

From DIN SPEC to an international standard

The project focuses on the development of the international technical specification ISO/IEC TS 25223, “Guidance and requirements for uncertainty quantification in AI systems.” Building on the national DIN SPEC 92005, the previous focus on machine learning is being expanded to cover AI systems as a whole – including hybrid and generative AI approaches. The goal is to define a uniform terminology as well as technical guidelines and requirements for uncertainty quantification that cover all phases of the AI system lifecycle.

In this way, UncertAInty makes a key contribution to international standardization and lays the groundwork for cross-industry standards. The results are incorporated into relevant international bodies and serve as an important reference for further standards in the fields of AI, functional safety, and systems engineering.

Legal certainty for high-risk AI

Grundlagen Cloud-Security - selbstbewusst und sicher mit Cloud-Technologien umgehen — © iStock.com/NicoElNino

A key driver of the project is the growing need for legal certainty in the use of AI, particularly in light of the European AI Act. The Act’s requirements leave room for interpretation, which must be clarified through harmonized standards. For companies, testing organizations, and regulatory authorities, quantifying uncertainty is therefore becoming a key issue.

Traditional safety standards are reaching their limits, particularly when it comes to safety-critical embedded AI systems, as they often do not account for AI components in the safety-critical path. UncertAInty addresses this gap and supports the development of standards that enable AI-based functions to be used safely without unnecessarily limiting their potential.

Fraunhofer IESE in the UncertAInty project

Reliable AI is a key focus area at Fraunhofer IESE, including a certified continuing education program. The institute has many years of expertise in quantifying and managing AI uncertainty gained from numerous research and industry projects and, through its preliminary work, has contributed significantly to the development of DIN SPEC 92005, for which it also served as deputy chair. This experience is incorporated into UncertAInty and supports the further development of internationally compatible standards.

Contribution to Trustworthy AI

UncertAInty lays the groundwork for the trustworthy use of AI. The project enhances the international compatibility of German standardization efforts and helps companies develop AI systems that are secure, traceable, and compliant. In doing so, UncertAInty makes an important contribution to competitiveness and the safe adoption of AI-based technologies in safety-critical applications.

More info: Add to the project website (attorney has been contacted)

Do you have a challenge for us?

Contact us!!

Are you looking to mitigate potential AI risks or determine whether risk management makes sense for your company?

Please schedule an appointment with us—we’d be happy to assist you via email or phone.

Reference Project: UncertAInty