Chapter V sets a separate set of duties for providers of general-purpose AI models. All such providers face the baseline duties in Article 53, providers established outside the Union face the authorised representative duty in Article 54, and providers of models with systemic risk face the additional duties in Article 55. Classification under Articles 51 and 52 decides which of those layers applies to a particular model.
What is a general-purpose AI model under the Act?
A general-purpose AI model is defined by reference to generality and capability. The Act describes a model that displays significant generality, is capable of competently performing a wide range of distinct tasks, and can be integrated into a variety of downstream systems or applications. That description turns on what the model can do across tasks, not on a single product feature or a marketing label. A narrow model built for one specific task is not automatically a general-purpose model because it uses a large amount of training compute. Conversely, a model that is presented as a component or an assistant can still be a general-purpose model if it meets the definition in the legal text. Teams should record the capabilities relied on, the tasks the model was designed to perform, and the evidence for that assessment, and should confirm the reading against the current consolidated text and Commission guidance before treating the label as settled.
How do Articles 51 to 56 fit together?
Chapter V works as a sequence. Article 51 sets the criteria for classifying a general-purpose AI model as presenting systemic risk, including a presumption linked to cumulative training compute and a route for Commission designation on other grounds. Article 52 then sets the notification, designation and reassessment procedure for models in that category. Article 53 sets baseline duties for all general-purpose AI model providers. Article 54 sets the authorised representative duty for providers established outside the Union. Article 55 adds duties that apply only to models with systemic risk. Article 56 provides for codes of practice that are intended to help providers apply Articles 53 and 55 in practice. Reading any one article alone gives an incomplete picture, because the classification step in Articles 51 and 52 determines whether Article 55 is engaged at all.
For the broader regulatory framework, see General-purpose AI models overview.
Why is the model layer separate from the system layer?
Chapter V duties attach to the model provider, while high-risk system requirements in Chapter III attach to the provider of a particular AI system. The same model can sit inside many downstream systems, each with its own intended purpose, affected people and deployment context. A downstream provider therefore cannot treat the model provider's Chapter V work as a substitute for its own system assessment, and a model provider cannot treat a downstream system's conformity work as discharging its own model duties. The boundary matters in contracts and in evidence: model documentation, downstream information and copyright policy answer model-level questions, while risk management, data governance and conformity assessment answer system-level questions. Teams should keep the two records distinct, link them where a system depends on a model, and record who is responsible for each layer.
When did Chapter V start to apply?
General-purpose AI model rules in Chapter V apply from 2 August 2025. That date is fixed in the verified date list and is distinct from the later application dates for high-risk systems. Providers of models placed on the market before 2 August 2025 are subject to a transitional rule in Article 111(3), under which they must take the necessary steps to comply by 2 August 2027. That transition is not a general exemption and it does not remove the need to plan the Article 53 and, where relevant, Article 55 work. Teams should record the date a model was placed on the market, any later change that could amount to a new placing, and the provision relied on for any transitional treatment. If the placing date or the effect of a later change is unclear, the question should be referred for qualified review rather than assumed in the provider's favour.
Who supervises general-purpose AI model providers?
Supervision of general-purpose AI model providers sits principally with the Commission, acting through the AI Office, rather than with a single national market surveillance authority for the model duties themselves. The AI Office has a role in classification, in receiving notifications and information, in facilitating codes of practice, and in enforcement steps set out in the Act for this layer. National authorities remain central to system-level market surveillance and to other parts of the Act. That split means a model provider and a downstream system provider may deal with different supervisors for different questions about the same product chain. Records should therefore identify which duty is being evidenced, which authority is relevant to that duty, and where a request or notification was sent. A single contact log that mixes model and system correspondence can obscure that distinction.
Continue with Article 51 classification of systemic-risk models.
What does this map not decide?
This map does not decide whether a particular model is a general-purpose AI model, whether it presents systemic risk, or whether any specific duty has been met. It does not set a training compute figure for a reader's model, assess a code of practice commitment, or determine the effect of a modification or a new version. Those questions turn on the model's capabilities, its training history, the consolidated legal text, and any designation or guidance that applies to that model. A provider should confirm its classification and its duties with a competent authority or legal adviser, keep the reasoning and the sources checked, and revisit the map when the model, its use, or the official guidance changes.
Frequently asked questions
Does Chapter V apply to every AI system?
No. Chapter V applies to providers of general-purpose AI models. A particular AI system built with such a model is assessed separately, including for high-risk classification and transparency duties. The two layers can both be relevant to one product, but neither replaces the other.
Do all general-purpose AI model providers face Article 55 duties?
No. Article 55 applies only to providers of general-purpose AI models classified as presenting systemic risk under Articles 51 and 52. All general-purpose AI model providers face the baseline duties in Article 53, subject to the limited open-source treatment described in that article, while Article 55 is an additional layer for the systemic-risk category.
Where should a provider start?
Start by recording the model, its capabilities, its placing-on-the-market date, and the classification reasoning under Article 51. Then map the Article 53 baseline, the Article 54 representative question if the provider is established outside the Union, and whether Article 55 is engaged. Keep each step linked to the consolidated text and to the date the check was made.
Continue through this AI Act cluster
This article is an informational overview, not a legal determination. Check the current legal text and official implementation guidance for your system and role.
Consolidated Regulation (EU) 2024/1689 ↗Commission AI Act Service Desk timeline ↗European Commission AI Act overview ↗