Description of the service

We audit AI models with respect to accuracy, robustness, fairness, explainability, security-related aspects and the plausibility of design decisions. The service is particularly relevant for manufacturing applications such as machine vision, human-robot interaction, autonomous production systems, predictive maintenance and AI-supported quality control. It supports preparation for the EU AI Act, especially Articles 13, 14 and 15 on transparency, human oversight, accuracy, robustness and cybersecurity, and is aligned with ISO/IEC 24029 and ISO/IEC 24027
Expected results:

The customer receives a structured audit report with a rating per criterion, identified weaknesses and prioritized recommendations for optimization and risk mitigation. For a Deep Audit, the report also includes own test results, metrics and quantified findings.

The methodology:

The service is available as an Assessment or as a Deep Audit. In the Assessment, the customer completes a structured questionnaire and we review the provided information for plausibility, completeness and consistency. This is suitable for an initial assessment, internal preparation or conformity planning and usually takes 2–4 weeks. In the Deep Audit, we receive access to the model via API or model artifact and perform independent technical tests. This is suitable for high-risk AI systems, external documentation or evidence towards customers, partners or authorities and usually takes 6–12 weeks.

Target:

Open for anyone

Improve your project
around production and
automation with AI technologies