Driver monitoring and assistance: which AI Act duties apply?

Short answer: AI functions for driver monitoring and driving assistance may fall under the AI Act's high-risk rules, but not via the standalone list in Annex III. They run through the Annex I route: AI as a safety component of a product that already falls under sectoral product legislation and type-approval. Whether and when the high-risk duties apply is therefore co-determined by that vehicle legislation.

Two routes to "high-risk"

The AI Act (Regulation (EU) 2024/1689) provides two ways for an AI system to qualify as high-risk. The first is Annex III, a standalone list of use cases. The second, in Article 6(1) and Annex I, applies to AI systems intended to be used as a safety component of a product covered by the Union harmonisation legislation listed there, or that are themselves such a product — provided the product requires third-party conformity assessment.

For vehicles, Annex I refers to the sectoral type-approval legislation. Driver monitoring and driving assistance embedded as a safety component in a vehicle can therefore, in principle, fall under the Annex I high-risk route. The exact scope follows from the interplay between the AI Act and the relevant sectoral legislation; consult the primary sources for this.

What this means in practice

If a system falls under the high-risk route, the requirements of Chapter III of the Regulation apply: risk management, data governance, technical documentation, event logging, transparency towards users, human oversight and appropriate accuracy, robustness and cybersecurity. For products with an existing conformity assessment, the Regulation aims to integrate these requirements as far as possible into the existing sectoral procedures, to limit duplicate assessment.

Timeline and points of attention

The dates of application differ per route. For high-risk systems covered by the Annex I product legislation, the Regulation sets a later date of application than for the Annex III systems. Because these dates are subject to ongoing amendment, it is advisable to verify the current state via the primary source and the Commission policy page before committing to a date.

Note that not every in-car AI is high-risk: functions that do not constitute a safety component, or products without a mandatory third-party assessment, may fall outside the high-risk category. A concrete classification always requires assessment against the Regulation's text and the sectoral legislation.

Read more: AI Act: timeline of obligations. Take the scan.