Technical Documentation
Detailed records demonstrating AI system compliance with regulatory requirements.
How the laws define this
Both sources agree that technical documentation consists of records and materials detailing system design, data sources or datasets, risk assessments and mitigation measures, and testing results that AI providers must prepare to demonstrate regulatory conformity. They diverge primarily in their practical focus and operational requirements. Regulation (EU) 2024/1689 — Artificial Intelligence Act establishes an ongoing lifecycle requirement, specifically mandating that providers keep these materials updated across the system's entire lifecycle and include logs and evidence of conformity assessments to enable market surveillance and audits. In contrast, the text for KI-Servicestelle (AI Service Office) — establishment at RTR focuses on compliance guidance, highlighting that the Servicestelle issues standardized templates and guidance to define what constitutes adequate technical documentation across different risk classes.
Synthesised from the 2 statutory definitions below. Each one is quoted in full, with its source.
Definitions (2)
Technical documentation denotes the set of documents and records (model design, datasets, performance, risk assessments, mitigation measures, testing results) that providers must prepare to demonstrate conformity with applicable AI regulatory requirements, as referenced in the AI Act and reflected in Servicestelle templates. The Servicestelle issues guidance and templates describing what constitutes adequate technical documentation for different risk classes.
The set of records and materials providers must prepare and keep up to date across an AI system's lifecycle, detailing system design, data sources and governance, risk management measures, testing and validation results, logs and evidence of conformity assessments to enable market surveillance and audits.
Related Terms
Conformity Assessment
Process to verify that high-risk AI systems meet regulatory requirements before market placement....
Risk Management System
Continuous process to identify, analyze, and mitigate AI system risks throughout its lifecycle....
Model Card
Standardized documentation providing key information about an AI model....