On May 7, 2026, a political agreement was reached in the trilogue on the EU “Digital Omnibus” package. The agreement concerns the so-called “Digital Omnibus on AI,” which is part of the EU “Digital Omnibus” package. The aim of the package is to simplify existing EU digital regulation and reduce regulatory burdens on businesses.
In the following article, we highlight the key changes and their implications for the automotive industry, particularly for the development and use of AI technologies in the automotive sector.
1. What changes does the Digital Omnibus on AI generally propose for the AI Act?
The so-called “Digital Omnibus on AI” aims to further simplify European digital regulation—particularly the Artificial Intelligence Act—reduce regulatory overlaps, and lessen the burden on businesses.
The political agreement reached so far focuses in particular on:
- clarifications regarding the classification of high-risk AI and postponement of the deadlines for the application of the regulations on high-risk AI,
- a narrower interpretation of the term “safety component,”
- simplifications regarding documentation and registration requirements,
- greater consideration of existing sectoral regulatory regimes,
- adjustments in the areas of AI governance and AI literacy,
- simplifications and clarifications regarding the use of special categories of personal data for bias detection and testing,
as well as an overall more risk-based and innovation-friendly approach.
The aim of the Omnibus is clearly not a complete overhaul of the AI Act, but rather an operational “fine-tuning” of the existing framework. This is particularly relevant for industries with already heavily regulated product and safety regimes—especially the automotive industry.
It is precisely in this sector that the question now arises as to what impact the changes will actually have on regulated vehicle products and in which areas the Omnibus will be practically relevant.
2. Will the Digital Omnibus on AI lead to fundamental changes for regulated vehicle products under Regulation (EU) 2018/858 and Regulation (EU) 2019/2144?
At first glance, the immediate impact of the Digital Omnibus on AI on traditional regulated vehicle products is likely to remain limited. The reason for this lies in the existing framework of the AI Act itself.
Already today, vehicles and certain vehicle components fall under the high-risk regime via Annex I, Part B of the AI Act, provided that:
- the respective products are subject to sectoral EU product regulation,
- a conformity assessment procedure is required, and
- the AI performs a safety function or constitutes a “safety component.”
For vehicles, however, regulatory integration is already taking place today and will continue to do so primarily through the respective Type Approval Framework regulations, such as
- Regulation (EU) 2018/858,
- Regulation (EU) 2019/2144,
- UNECE regulations such as R155 and R156,
as well as functional safety and type approval processes.
The Digital Omnibus does not currently appear to intend to fundamentally change this basic sectoral framework.
3. Does the Digital Omnibus on AI nevertheless create new uncertainties or changes for regulated automotive AI?
Yes — despite the limited immediate changes, several important new questions regarding interpretation and structure arise.
Particularly relevant is the planned clarification of the term “safety component.” This could narrow the scope of vehicle functions actually classified as high-risk AI in the future.
Until now, there has been considerable uncertainty as to whether virtually any AI-supported vehicle function that influences vehicle operation could be classified as high-risk AI. The Omnibus now points toward a narrower and more safety-focused interpretation:
Systems that exclusively concern comfort, performance, service efficiency, or quality control, and whose malfunction does not pose a risk to health or safety, should not be classified as safety components and thus not automatically classified as high-risk.
This could allow for a more nuanced distinction in the future between:
Vehicle functions that remain relatively high-risk:
- ADAS and automated driving functions,
- driver monitoring systems,
- AI-supported braking, steering, or stability functions,
- safety-relevant sensor fusion and object detection,
and
AI functions that are generally not high-risk:
- comfort and personalization functions,
- infotainment AI,
- Pure voice assistants,
- range or efficiency optimization,
- predictive maintenance without direct safety implications,
- Backend analysis and development systems.
The focus is thus increasingly shifting to the question of whether a specific AI is actually part of a safety-relevant product function.
For OEMs and suppliers, robust classification and demarcation analysis is therefore becoming particularly important.
This clarification is likely to lead to a closer alignment between AI Act assessments and established automotive safety concepts — such as those defined by ISO 26262, ISO 21448 (SOTIF), and UNECE Regulations R155 and R156. For OEMs and suppliers, this opens up the possibility of aligning high-risk assessments more closely with existing safety-related hazard and risk assessments in the future, rather than considering each AI component in isolation from a regulatory perspective.
4. What impact does the Omnibus have on the use of special categories of personal data in the automotive sector?
The Omnibus also explicitly addresses the use of special categories of personal data—such as health data or biometric data—for purposes such as bias detection, bias correction, or AI testing.
However, in the automotive sector, the fundamental question that arises first is how and where these regulations are relevant at all.
The discussion is particularly interesting because the planned new regulations are to be enshrined horizontally within the AI Act itself, while the practical regulation of vehicle AI for Annex I Part B products is largely carried out through sectoral product regimes.
This raises a systemic question:
Articles 102 et seq. of the AI Act do usually not provide that all provisions of the AI Act must automatically be implemented or taken into account in sectoral product regimes. Rather, the regulation refers only to certain requirements. For regulated technologies in the automotive sector, these specifically include the requirements under Chapter III, Section 2 of the AI Act (requirements for high-risk AI systems), but not, for example, the additional provisions in Section III (obligations of providers and operators of high-risk AI systems and other stakeholders). If the new regulations on the use of special categories of personal data are not implemented in Articles 8–16 of the AI Act, they would not be covered by the mandate of Articles 102 et seq. of the AI Act (consideration in the design of sectoral requirements).
As a result, it is not yet entirely clear:
- whether and how the new bias detection regulations will practically “trickle down” to vehicle products,
- whether sectoral implementing acts or guidance will be required, or
- whether the horizontal AI Act regulation is sufficient on its own.
The concrete implementation remains to be seen in this regard. However, with the proposed new regulations, the EU legislator has at least provided an “impulse” that will simplify the processing of relevant data, including in the areas of automotive development and testing.
In practice, relevance in the automotive sector is likely to arise only in certain special cases anyway, particularly with:
- driver monitoring systems,
- biometric driver identification,
- interior camera systems,
- fatigue and attention detection, or
- certain personalized in-cabin and wellness features.
For traditional vehicle AI such as environmental sensors, driving trajectories, sensor fusion, or predictive maintenance, however, the processing of special categories of personal data is unlikely to play a central role on a regular basis.
5. For which automotive products are the Omnibus amendments likely to be most relevant?
The main practical impact of the Omnibus on AI is likely to be felt outside the scope of traditional regulated vehicle products.
This is because numerous AI systems in the automotive sector do not fall under the sector-specific special regimes established by Regulation 2018/858 or Regulation 2019/2144.
This applies in particular to:
- Development and engineering AI,
- AI-supported quality control,
- Production and manufacturing AI,
- Supply chain and logistics systems,
- Recruiting and HR AI,
- Backend analytics,
- Dealer and customer platforms,
- Voice and assistant systems outside of safety-critical vehicle functions,
- as well as internal AI governance and enterprise systems.
For such systems, the changes introduced by the Omnibus are likely to be significantly more directly relevant, particularly with regard to:
- High-risk classification,
- Documentation requirements,
- Registration requirements,
- AI literacy,
- Data governance,
- as well as potential exemptions for certain AI categories.
It is precisely here that the Omnibus regulation is likely to have its greatest practical impact in terms of relief.
The extended transition periods of 12 or 24 months (“stop-the-clock”) are more than welcome for the automotive industry, which is already overwhelmed by regulatory requirements.
The deadlines for applying the obligations for high-risk AI systems are linked to the availability of standards, common specifications, or guidelines. Once these standards are in place, there will be a 12-month transition period for AI systems integrated into regulated products (Annex I). However, the aforementioned long-stop dates (December 2027 for Annex III systems / August 2028 for Annex I systems) are in any case considered absolute deadlines.
6. What should OEMs and suppliers do now?
Where applicable, OEMs and suppliers should not view the “gained” transition periods as a “break,” but rather as an “extended” preparation window.
In addition, OEMs and suppliers should not view the bus solely from the perspective of traditional vehicle products. Rather, a clearer distinction will be required in the future between:
Regulated vehicle AI systems with a focus on product safety, type approval, functional safety, and safety components, and other AI systems within the company, which may fall directly under the general AI Act regime.
The following steps are therefore particularly recommended:
- Updating existing AI Act gap analyses and project timelines,
- Mapping all AI systems across the vehicle and corporate lifecycle,
- Differentiated classification of regulated and unregulated AI,
- Establishing robust safety component and high-risk assessments,
- Integration of AI governance into existing safety, cybersecurity, and homologation processes, as well as
- Monitoring further sector-specific guidance and implementation acts.
Suppliers in particular should review
- what evidence OEMs will contractually require in the future,
- what documentation and audit obligations will arise along the supply chain, and
- how existing development processes can be harmonized with AI Act compliance.
Of particular relevance here is the integration with existing standards such as UNECE Regulation No. 155/156, ISO/SAE 21434, ISO 21448 (SOTIF), ISO 26262, and ISO 42001 (AI Management System – AIMS).
In particular, the distinction between regulated vehicle AI and other corporate AI is likely to become one of the central governance challenges for the automotive industry in the future.
7. What is the timeline moving forward—when will the changes take effect?
Following the political agreement in the trilogue, the formal steps now remain:
- formal adoption by the Parliament and the Council,
- publication in the Official Journal of the EU,
- followed by the entry into force of the amending regulation.
As things stand, the rules for high-risk AI in particular are set to be postponed:
- for standalone high-risk AI, likely by December 2027 at the latest,
- for AI systems in regulated products by August 2028 at the latest.
The final versions of the amendments and specific transitional provisions are still pending.
For the automotive industry in particular, the final structure of the interplay between the AI Act, type-approval law, and sector-specific safety regimes—which is currently still under development and will require close monitoring—is likely to be decisive. The process should be closely monitored to ensure a timely response to future requirements.
