General-purpose AI models versus high-risk AI systems: what is the difference?

The EU AI Act treats a general-purpose AI model and a high-risk AI system as two different things. A general-purpose model, such as a large language model, is regulated under Chapter V by its provider, with extra duties where it has systemic risk. A high-risk AI system is regulated under Chapter III by its provider and its deployer, because of what it is used for.

A company can be a provider of a general-purpose model and a provider or deployer of a high-risk system at the same time, and the two sets of duties stack. The general-purpose model duties applied from 2 August 2025; the high-risk duties apply from 2 December 2027 or 2 August 2028, depending on the route.

General-purpose AI models versus high-risk AI systems: what is the difference: the source

Read the regulation on EUR-Lex and the European Commission AI Act Service Desk, both linked from every obligation page. Application dates are Article 113, as amended by Regulation (EU) 2026/1744.

Where to go next

The full 43-obligation directory, or the conformance kit for per-role checklists and the evidence register.