The EU AI Act in the agrifood sector: what it requires, of whom, and when

The EU AI Act does not automatically make every use of artificial intelligence in food and agriculture a high-risk system. For an agrifood business, classification depends less on the technological label than on the specific function performed by the system: whether it operates as a safety component of regulated machinery, whether it affects employment-related decisions, or whether it remains an operational support tool in planning, inventory, quality control or customer service.

This report summarises the obligations already in force, the requirements that will apply between 2026 and 2028, and the points at which the sector most often misreads the Regulation. The practical conclusion is deliberately cautious: the postponement of high-risk obligations gives companies time to organise inventories, evidence and controls, but it does not remove the need to build a minimum audit trail for AI systems already in use or under consideration.

As of 1 June 2026, prohibitions and AI literacy requirements have applied since February 2025; obligations for general-purpose AI models have applied since August 2025; Article 50 transparency duties start in August 2026; and, under the political agreement on the AI Omnibus, high-risk obligations move to December 2027 for stand-alone Annex III systems and to August 2028 for systems embedded in regulated products.

In agrifood, the central distinction is between food safety, machinery safety and workforce management. A vision system that classifies batches may be a decision-support tool; the same system, if it stops a line when a human hand enters a cutting zone, may become a safety component. If it is also used to assess individual performance or inform contract renewal, it enters the labour-related territory of Annex III. Three functions, three regulatory regimes, one piece of equipment.

The operational recommendation is to begin with a functional inventory: which systems exist, which decisions they support or automate, which data they use, who supervises their outputs, and what evidence is retained. For small and medium-sized agrifood companies, the advantage will not come from over-compliance, but from disciplined documentation: explaining why a system is not high-risk where appropriate, integrating controls into HACCP, ISO 22000 or ISO 9001 when relevant, and asking providers for intelligible documentation on data, logs, human oversight, robustness and cybersecurity.

The full report can be read on Google Drive:

https://drive.google.com/file/d/1w8bCoYONMhTSZojoAwLwxrLYVXTMx4fK/view?usp=drive_link

This text is intended as strategic and informational analysis; it does not constitute legal advice.

Main sources: European Commission, AI Act; Council of the EU, AI Omnibus agreement of 7 May 2026; Annex III of the AI Act.

Research by: Beatriz Vallina, PhD

Thesis Supervisors: Roberto Cervelló, Prof.PhD & Juan José Lull, PhD

Institution: Doctorate in Agrifood Economics, Universitat Politècnica de València

© 2025. All rights reserved.