TM2.0 has submitted its response to the European Commission’s consultation on the draft Guidelines for classifying high-risk AI systems under Article 6 of the AI Act. Filed on 23 July 2026, the submission draws on input from TM2.0 members across the public and private sectors and sets out how AI-enabled systems in road traffic management should be treated under the new rules.
The response is based on the Platform’s Open Letter from May 2026, which was signed by 28 supporting organisations, calling for structured dialogue with the sector and cautioning against a one size fits all approach to high-risk classification.
Traffic management sits at the crossroads of critical infrastructure, public service delivery and increasingly advanced digital systems. AI already supports forecasting, optimisation, monitoring and operational decisions across the sector, but not every AI function carries the same level of risk, and the Guidelines need to reflect that. With the relevant obligations set to apply from 2 December 2027 under the AI Omnibus timeline, the sector has limited time to get the framework right. TM2.0’s central argument is that an AI system should be judged by its intended purpose, its actual function and its role in the wider architecture, not by the importance of the infrastructure around it or by hypothetical downstream effects.
A key recommendation is to separate AI systems that directly perform, control or supervise a safety function from those that simply support operations, things like traffic-flow forecasting, congestion prediction, simulation, travel-time estimation, network monitoring and decision support. These tools improve the quality of traffic decisions, but TM2.0 believes that they shouldn’t be classified in the same way as systems that issue safety-relevant commands or determine the state of critical infrastructure. The test should be whether the AI can directly create, prevent, control or mitigate an unsafe physical condition, not whether it affects congestion, delays or general performance.
The submission also points to how AI interacts with certified controllers and validated safety mechanisms already used across the sector. In many systems, AI optimises performance while separate, validated components keep enforcing safety. TM2.0 calls these “bounded AI systems,” since they operate within fixed limits and can’t override the safety layer. Traffic signal control provides a useful example: an AI layer may support optimisation and operational efficiency, while separate validated and certified mechanisms remain responsible for preventing unsafe signal states. In such cases, the AI layer should not automatically be treated as the safety component simply because it supports the operation of the overall system.
TM2.0 therefore calls for clearer guidance on how classification should interact with existing standards and certification processes.
Because modern traffic systems combine many components and data sources, TM2.0 also recommends that classification reflects the actual role of the AI within the full solution, rather than extending high-risk status to every supporting part just because it feeds into a larger system. The combined system should only be assessed together where the integration itself creates the safety function, rather than automatically extending classification to every supporting component within the wider architecture.
TM2.0 also makes one practical request: a clearer decision framework distinguishing classification under Article 6(2) from the filtering mechanism under Article 6(3). Such a framework would help stakeholders assess whether an AI system performs a safety function or merely supports one, whether it operates alongside validated non-AI safety mechanisms or presents a real safety risk.
TM2.0 is also asking for more real-world examples in the final Guidelines to improve legal certainty and illustrate how the distinction between operational support functions and safety-relevant AI systems should be applied in practice. Such examples should show clearly where optimisation ends and safety control begins.
TM2.0 remains ready to support the European Commission and the AI Office in shaping guidance that protects safety, gives the sector legal certainty and allows responsible AI innovation to move forward across Europe’s transport system.
