AI regulation in logistics and transport: four routes that hit the sector

Anyone in logistics or transport who "needs to do something with the AI Act" usually gets a generic story about high-risk systems and fines. That helps no one: the sector is hit along four different routes, on four different timelines — and the most urgent one is not on the horizon but a prohibition that already applies.

Route 1 — Driver and worker monitoring: the ban that applies now

Since 2 February 2025, Article 5 prohibits emotion recognition in the workplace. That hits the sector directly: systems inferring the mood, stress or emotions of drivers, order pickers or planners — from camera footage, voice or behaviour — are banned, save for a narrow exception for medical or safety purposes. Fatigue detection demonstrably aimed at safety can fall within that exception; monitoring that in fact measures performance or attitude cannot. The guidelines of 4 February 2025 draw that line deliberately tight. Anyone running fleet or warehouse monitoring should know today on which side of the line each system sits.

Route 2 — Planning, rostering and recruitment: high-risk (Annex III, point 4)

AI that makes or steers decisions on recruitment, task allocation, promotion or termination, or that monitors and evaluates worker performance and behaviour, is high-risk. In logistics that is not a side-show but the core of the operation: rostering and planning algorithms assigning trips and shifts, per-driver or per-picker performance dashboards, recruitment filters for scarce staff. The obligations (risk management, representative data, human oversight, logging, informing workers) are expected to apply from 2 December 2027 — but the systems you procure now will still be running then.

Route 3 — Warehouse robots and vehicles: the machinery side (Annex I)

AGVs, autonomous mobile robots and other smart machinery fall under the new Machinery Regulation (EU) 2023/1230, applicable from January 2027 — and that regulation is listed in Annex I of the AI Act. The consequence: an AI system performing the safety function of such a machine is automatically high-risk, with conformity assessment through the machinery route (AI Act requirements expected from 2 August 2028). For warehouse automation that means a double supplier question: does the machine comply with 2023/1230, and does the AI component comply with the AI Act?

Route 4 — Traffic and infrastructure (Annex III, point 2)

AI as a safety component in the management of road traffic and critical infrastructure is its own high-risk category. Traffic management, corridor control and port logistics that lean on AI for safety decisions fall here — relevant for road authorities, ports and the larger chain orchestrators.

And running through everything: the duties that already apply

Apart from the four routes, every company in the chain faces the AI literacy obligation (Article 4, since February 2025) for everyone working with AI — from planner to management — and soon the transparency duties of Article 50 for customer-facing chatbots and AI content (expected from 2 December 2026). Whoever uses generative AI in planning or customer service is already inside the regime.

The practical order

Three steps that can start this week: map per site which systems fall into which route (the AI register); put the two hard questions to suppliers of monitoring and planning systems (which route, which date, whose conformity); and sort Article 4 — the only obligation with no deferral clause whatsoever. This knowledge base tracks all four routes; the timeline and the high-risk overview are the adjacent files.