As the European Parliament moves toward finalising its position on the Digital Omnibus on AI, a wave of plenary amendments has highlighted a central tension: how to simplify the AI Act without stifling European competitiveness or compromising safety.
Following the Joint Committee (LIBE/IMCO) report on 18 March, various political groups have tabled amendments that seek to redefine the "rules of engagement" for high-risk AI, particularly for startups and SMEs within the smart mobility and transport sectors.
Innovation vs. Regulation: The Political Battle Lines
Several key amendments tabled ahead of the 26 March plenary vote aim to protect the EU’s industrial logic:
- Protecting SMEs and startups (PfE & ECR): Shadow Rapporteurs from the PfE (Virginie Joron, Jaroslav Bžoch) and ECR (Assita Kanko, Piotr Müller) have introduced points emphasising that legislation must not impede AI investment. The ECR specifically proposed exempting internal AI systems used for essential business operations that pose minimal risk, aiming to slash "unnecessary bureaucracy."
- Revised compliance deadlines: There is a significant push to push back the clock for high-risk systems. Both PfE and ECR have proposed moving the application date for Chapter III (High-risk AI) to 2 August 2028 for systems classified under Annex III, providing a longer runway for high-investment industrial systems.
- Bias detection vs data privacy (GUE/NGL & ESN): Amendments from the GUE/NGL and ESN groups clarify that while providers may process sensitive personal data for bias detection and correction, this must be done under strict necessity. ESN Rapporteur Mary Khan emphasised that such data must be pseudonymized and strictly segregated from general model training.
Key Technical Amendments for the Mobility Sector
For ERTICO partners, two specific procedural amendments are of particular note:
- Source inspections: New points (26a and 26b) suggested by the ECR would permit on-site inspections of source code for cybersecurity purposes, while ensuring all supervisory communications are systematically recorded to enhance transparency.
- Child safety and "nudifiers": The PfE group proposed a new prohibition on AI systems that manipulate child sexual abuse material, aligning the Omnibus with broader European digital safety mandates.
The Road to August 2026
The proposal follows the ordinary legislative procedure. With the Council having adopted its mandate on 13 March and the Parliament's plenary vote finalised on 26 March, trilogue negotiations are now underway.
The goal is to reach a first-reading agreement by May 2026, allowing for publication in the Official Journal by July and entry into force in August 2026, just as the general application of the AI Act begins.
Source: European Parliament (24 March 2026): Artificial intelligence: Parliament to vote on nudification ban