Berlin – September 09, 2026 -- Germany's TÜV testing organizations are building a joint certification framework for artificial intelligence systems, with market rollout targeted for late 2026, as new survey data shows 79 percent of German citizens fear uncalculable risks from AI.
Survey shows trust gap even among active AI users
According to the TÜV Association's 2025 AI study, a representative survey conducted around October 26, 2025, 79 percent of German citizens fear unpredictable AI risks. Among people who actively use AI applications, only 53 percent trust that the results and answers produced are correct.
TÜV Association coordinates certification framework with pilot partners
The TÜV Association is coordinating development of the framework, with TÜV AI.Lab building the technical requirements and test catalogs, and TÜV company experts contributing certification and real-world AI assessment experience. The certifications are currently being tested with pilot partners under live operating conditions, with findings feeding directly into further development of the program before availability to companies from the end of 2026.
Framework offers three certification tiers matched to risk level
The program comprises three distinct AI certifications differing in scope, depth and methodology. The first targets AI systems that fall outside the AI Act's high-risk category and checks baseline quality and safety requirements. The second is open to all AI systems and includes a structured, AI-specific risk assessment based on documentation review and on-site manufacturer audits. The third adds independent technical testing of selected system properties to allow deeper evaluation of performance and reliability. Dr. Christoph Poetsch, Head of AI Quality and Ethics at TÜV AI.Lab, said the framework was deliberately structured so that AI systems are tested according to the certification tier matching their specific use case, whether standalone applications or AI embedded in technical products.
Certification is designed to complement, not replace, EU AI Act compliance
The framework is built to be fully compatible with the EU AI Act and adds risk-based and technical testing aspects on top of the regulation's requirements. Marc Fliehe, head of digitalization and AI at the TÜV Association, said companies developing or deploying AI need to demonstrate that their systems function reliably and that risks are controlled, arguing that independent certification could become an important quality marker in the AI market. AI systems are already deployed in industrial production processes, medical diagnostics, driver assistance systems and recruiting software, where errors can carry significant consequences for people, companies or technical processes.
Rollout timeline confirmed for end of 2026
The phased introduction of the AI certifications remains scheduled for the end of this year, with the pilot phase continuing to validate testing requirements and procedures under real deployment conditions.