Guidelines on technical documentation under the EU AI Act

Antonio Tejeda Encinas, President of the Euro-American Committee of Digital Law - CEA Digital Law

The recent adoption of the Regulation of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (AI) in the European Union marks a new milestone in the legal regulation of these systems. The purpose of this Artificial Intelligence Act (AI Act) is to improve the functioning of the internal market through a uniform legal framework that encourages the development, placing on the market and use of AI systems in line with the Union’s values. It promotes human-centric, trustworthy and safe AI while ensuring the protection of health, safety and fundamental rights.

The AI Act establishes a harmonised legal framework for the development, marketing and use of AI products and services in the European Union. Its specific objectives include:

1. Ensuring that AI systems placed on the EU market are safe and comply with existing law.
2. Providing legal certainty to facilitate investment and innovation in AI.
3. Improving governance and the effective enforcement of EU law on fundamental rights and of the safety requirements applicable to AI systems.
4. Facilitating the development of a single market for lawful, safe and trustworthy AI applications and preventing market fragmentation.

A key obligation for providers of high-risk AI systems is to draw up comprehensive technical documentation before the system is placed on the market. This documentation must demonstrate compliance with the requirements of the AI Act and make it easier for the regulatory authorities to carry out their assessment, while providing legal certainty for deployers.

Article 11 of the AI Act requires this documentation to be drawn up and kept up to date throughout the life cycle of the AI system. Its purpose is to provide clear and comprehensive information to the national competent authorities and to deployers.

The technical documentation must include:

a) A general description of the AI system, its intended purpose, its interaction with hardware and software, the user interface and the instructions for use.
b) Details of the system’s development process, including methodologies, design specifications, system architecture, third-party components, the data sets used, human oversight measures, and validation and testing procedures.
c) Information on the monitoring, functioning and control of the system, including its capabilities and limitations, unforeseen outcomes, human oversight measures and input data specifications.
d) The performance metrics of the AI system.
e) A description of the risk management system.
f) A record of the changes made to the system throughout its life cycle.
g) A list of the harmonised standards applied.
h) A copy of the EU declaration of conformity.
i) A description of the post-market evaluation system, including the monitoring plan.

This documentation must be kept for ten years from the date on which the high-risk AI system is placed on the market or put into service. It must also be consistent with the manufacturer’s AI governance policy and with the documentation relating to CE marking.

The technical documentation must be thorough enough to demonstrate compliance with the AI Act and to allow effective monitoring of operations and post-market control. For high-risk AI systems related to products covered by EU harmonisation legislation, a single set of technical documentation may be drawn up, combining the information required by the AI Act and by the relevant legislation, so as to minimise the additional burden on providers.