In the run-up to the plenary meeting, various working groups had already met to further develop key standards. These included work on the maintenance and further development of TTCN-3 (TTF 056), the globally established test and specification language for interoperable and automated testing, as well as the activities of the MTS_AI working group on the standardisation of testing, auditing and documentation procedures for AI systems (TTF 058).

Fraunhofer FOKUS has been actively involved in the ETSI MTS Committee for many years and chairs the MTS_AI working group. The aim is to develop methods and standards that support the safe, trustworthy and regulatory-compliant use of AI systems.

Among the most significant outcomes of recent years are three ETSI specifications that have been significantly shaped by the work of MTS_AI:

• ETSI TR 103 910 – AI Testing: Test Methodology and Test Specification for ML-based Systems describes test methods, quality criteria and procedures for evaluating machine learning systems. Among other things, the report addresses robustness, bias, security and the integration of testing throughout the entire ML lifecycle.
• ETSI TR 104 119 – Guidelines for Documentation of AI-enabled Systems provides comprehensive guidance on the structured and continuous documentation of AI systems. The specification addresses various stakeholders and forms of documentation, including model cards, fact sheets and other approaches to support transparency and AI Act compliance.
• ETSI TS 104 008 – Continuous Auditing Based Conformity Assessment (CABCA) for AI-enabled Systems defines a novel procedure for the continuous conformity assessment of AI systems. Instead of one-off audits, CABCA relies on the ongoing collection of evidence, automated checks and continuous assessment throughout the entire operation of an AI system. This creates an important foundation for the monitoring of dynamic and learning systems.

The current work of MTS_AI focuses in particular on the practical implementation of the requirements of the European AI Act. To this end, documentation schemes, auditing procedures and test methods are being developed that both meet regulatory requirements and are applicable in industrial development processes. ETSI TC MTS is thus positioning itself as one of the leading European standardisation bodies for quality assurance, auditing and documentation of AI systems.

The ETSI MTS#98 meeting at Fraunhofer FOKUS underscores the growing importance of standardised procedures for the development of trustworthy AI. The specifications TR 103 910, TR 104 119 and TS 104 008 provide concrete building blocks for testability, transparency and continuous conformity assessment – and thus for the practical implementation of trustworthy and AI Act-compliant AI in Europe.