EU AI Act Annex IV Technical Documentation Guide
Annex IV is the EU AI Act's prescribed contents list for the technical documentation that every provider of a high-risk AI system must compile and keep up to date.
What Annex IV technical documentation is
Annex IV technical documentation is the structured evidence file, defined by Article 11 and detailed in Annex IV of Regulation (EU) 2024/1689, that demonstrates a high-risk AI system meets the EU AI Act's requirements. It is the single dossier a provider relies on to show conformity, and it is what a notified body or market surveillance authority will ask to see.
Article 11 requires the documentation to be drawn up before the system is placed on the market or put into service and to be kept current throughout the system's lifetime. It must be clear and comprehensive enough for an authority to assess whether the system complies. The minimum contents are fixed by Annex IV, so a provider cannot simply decide which topics to cover.
Who needs it
The obligation falls on the provider of a high-risk AI system, meaning the organisation that develops the system or has it developed and places it on the EU market or puts it into service under its own name or trademark. High-risk covers AI used as a safety component of regulated products and the use cases listed in Annex III, such as employment and worker management, access to essential services, credit scoring, biometrics, and education.
Most high-risk obligations, including the Article 11 documentation duty, apply from 2 August 2026, with high-risk AI embedded in products already regulated under existing EU law following from 2 August 2027. Deployers do not draw up Annex IV documentation, but they should request and retain relevant information from the provider to meet their own obligations.
What Annex IV must contain
Annex IV sets out the topics the file must address. In practice they fall into six areas:
- System description: what the system is, its intended purpose, the provider, versions, how it interacts with hardware or other software, and the deployment forms, for example software on a device or a cloud service.
- Development process: the design choices, system architecture, data requirements and the datasets used for training, validation and testing, plus the human oversight measures built in.
- Performance metrics: the accuracy, robustness and cybersecurity levels the system is designed for, the metrics used to measure them, and the foreseeable limitations and trade-offs.
- Risk management: the risk management system required by Article 9, including the risks identified, the mitigations applied, and any residual risks that remain.
- Post-market monitoring: the system in place to collect and review performance once the system is in use, in line with Article 72, so issues can be detected and corrected.
- Changes and validation: a record of changes made through the lifecycle, the validation and testing procedures and their results, and the EU declaration of conformity together with applied harmonised standards.
Simplified documentation for SMEs
The EU AI Act recognises that small and medium-sized enterprises, including start-ups, cannot carry the same documentation burden as large providers. Article 11 directs the Commission to provide a simplified technical documentation form aimed at the needs of SMEs and small companies.
An SME that is a high-risk provider may use this simplified form, supplying the Annex IV elements in the equivalent simplified format. The substance still has to be there: the simplified form reduces overhead and presentation, not the underlying obligation to demonstrate conformity. A notified body involved in conformity assessment can still request the full set of elements where needed.
How ComplyAgent helps
ComplyAgent auto-drafts your Annex IV technical documentation from the answers you give during classification, mapping each section of the file to the Article 11 and Annex IV requirements so nothing is missed. Because you classify each AI system once, the same evidence also feeds your ISO/IEC 42001 and NIST AI RMF work, instead of being recreated three times.
The draft stays linked to your AI system inventory and risk records, so when a system changes the documentation can be updated rather than rewritten, keeping the file current as Article 11 requires. SME users can produce documentation aligned with the simplified approach without losing the structure an auditor expects.
Related
- How to classify an AI system
- Is my AI system high-risk?
- EU AI Act compliance software
- High-risk AI system (glossary)
Get started
Start your free trial with ComplyAgent, or see pricing.
Last reviewed June 2026 by the ComplyAgent team.