TÜV Organisations Develop Certification for Trustworthy AI

Berlin, 7 September 2026 – Artificial intelligence is rapidly becoming part of business and everyday life. At the same time, reports of incorrect outputs, bias and other risks associated with AI systems are increasing. This is affecting trust in the technology: According to a study by the TÜV Association, 79 per cent of people in Germany fear unpredictable risks from AI. Even among users of AI applications, only just over half (53 per cent) trust that the outputs and responses they receive are accurate.

“Trust is becoming a decisive factor in the further adoption of artificial intelligence,” says Marc Fliehe, Director of Digitalisation and AI at the TÜV Association. “Companies need to be able to demonstrate to customers, business partners and end users that their AI systems are not only capable, but also operate reliably and safely.”

The TÜV organisations are therefore developing a joint framework for the certification of AI systems. In future, manufacturers and providers will be able to obtain independent verification that their systems meet defined quality and safety requirements. The AI certification schemes are currently being piloted with partners under real-world conditions and are scheduled to be launched from late 2026.

“Artificial intelligence will only achieve lasting adoption if people and businesses can trust it,” says Fliehe. “Anyone developing or deploying AI needs to demonstrate that their systems operate reliably and that risks are properly managed. The TÜV AI certification schemes provide an independent and transparent means of doing so. It has the potential to become an important mark of quality in the AI market.”

Independent Assessment Provides Guidance in the AI Market

AI systems are already being used in industrial production processes, medical diagnostics, driver assistance systems and recruitment software. Depending on the application, errors can have significant consequences for people, businesses or technical processes. Artificial intelligence also has specific characteristics that need to be taken into account: outputs may, for example, systematically disadvantage certain groups, or systems may behave unreliably in unexpected situations.

The EU AI Act establishes a binding regulatory framework for artificial intelligence in Europe. At the same time, companies increasingly need independent assurance of the quality and safety of their AI systems. “A TÜV certificate can provide customers and business partners with an additional point of reference and help companies build trust in their AI applications,” says Fliehe.

Pilot Phase Underway – Market Launch in Late 2026

The framework and the individual AI certification schemes are being developed jointly by the TÜV Association, TÜV AI.Lab and the TÜV organisations. The TÜV Association is coordinating the project. TÜV AI.Lab is developing the technical and methodological foundations, including assessment requirements and detailed test criteria. Experts from the TÜV companies are contributing their experience in testing and certification as well as in the practical assessment of AI systems.

The TÜV organisations are currently piloting the AI certification schemes with partners. The pilot phase is examining how the assessment requirements and procedures perform under real-world operating conditions. The findings are being incorporated directly into the further development of the programme.

Three AI Certification Schemes Tailored to Different AI Systems

The framework comprises three AI certification schemes that differ in the scope, depth and methodology of the assessment. The appropriate certification scheme depends on the specific use case and the level of assurance required. The framework is fully compatible with the requirements of the EU AI Act and complements them with targeted risk-based and technical assessment elements.

The first AI certification scheme is designed for AI systems that are not classified as high-risk under the AI Act. It assesses fundamental quality and safety requirements. The second AI certification scheme is available for all AI systems. It includes a structured, AI-specific risk assessment and examines how the relevant requirements have been implemented, based on a review of the documentation as well as an on-site assessment at the manufacturer’s premises. For the third AI certification scheme, selected characteristics of the AI system are additionally assessed through independent technical testing. This enables a more in-depth evaluation of the system’s performance and reliability.

“Not every AI system needs to be assessed in the same way,” says Dr Christoph Poetsch, Head of AI Quality and Ethics at TÜV AI.Lab. “That is why we deliberately designed the AI certification framework to offer different AI certification schemes. Whether it is a standalone AI application or AI integrated into a technical product, manufacturers’ AI systems are assessed in a manner proportionate to risk under the AI certification scheme that best fits their specific use case.”

The phased introduction of the AI certification schemes is planned for late 2026. Through these AI certification schemes, the TÜV organisations are applying their testing and assessment expertise to independently evaluate the quality and safety of AI systems according to the specific use case and the required depth of assessment.

Methodology Note
The figures on public trust in artificial intelligence are taken from the TÜV AI Study 2025. The nationally representative survey was conducted on 26 October 2025.