Skip to main content

Article 53 for Autonomous Systems: Obligations of General-Purpose AI Model Providers

How Article 53 (obligations of providers of general-purpose AI models) concerns autonomous systems: who is bound, what is documented, the open-source exception, dates and fines under Article 101.

Providers of general-purpose AI models must document the model, inform downstream providers, publish a training-data summary and keep a copyright policy (Article 53(1)).

A vehicle, aircraft or robot maker that only integrates or uses a model is not bound by Article 53. The duties apply since 2 August 2025; models placed on the market earlier have until 2 August 2027 (Article 111(3)). Since 2 August 2026 the Commission can fine providers up to 3% of worldwide annual turnover or EUR 15 million, whichever is higher (Article 101). Next step: find who provides each model you use.

Article 53(1) sets four duties for model providers

These duties apply to every general-purpose AI model, whatever the sector of use.

  • Art. 53(1)(a)
    Technical documentation

    Draw up and keep up to date the technical documentation of the model, including its training and testing process and evaluation results, with at least the information in Annex XI, to provide on request to the AI Office and the national competent authorities.

  • Art. 53(1)(b)
    Information for downstream providers

    Draw up, keep up to date and make available information and documentation to providers of AI systems that intend to integrate the model, with at least the elements in Annex XII. It must let them understand the model’s capabilities and limitations and comply with their own obligations. This is without prejudice to the need to observe and protect intellectual property rights and confidential business information or trade secrets under Union and national law.

  • Art. 53(1)(c)
    Copyright policy

    Put in place a policy to comply with Union copyright law, in particular to identify and comply with a reservation of rights expressed under Article 4(3) of Directive (EU) 2019/790.

  • Art. 53(1)(d)
    Training-content summary

    Draw up and make public a sufficiently detailed summary of the content used for training, according to a template provided by the AI Office.

  • Art. 53(2)
    Open-source exception

    The duties in points (a) and (b) do not apply to models released under a free and open-source licence that allows access, usage, modification and distribution, and whose parameters, architecture information and usage information are public. The exception does not apply to models with systemic risk (Article 53(2)).

  • Art. 53(3)
    Cooperation

    Providers must cooperate as necessary with the Commission and the national competent authorities.

  • Art. 53(4)
    Codes of practice

    Providers may rely on codes of practice within the meaning of Article 56 to demonstrate compliance until a harmonised standard is published. Providers that follow neither must demonstrate alternative adequate means of compliance for assessment by the Commission.

Who in autonomous systems is bound by Article 53

Article 53 binds the provider of the model, whatever the sector or end product.

  • Autonomous systems is not a category in the Act

    Whether a model is a general-purpose AI model depends on the definition in Article 3(63): significant generality, competent performance of a wide range of distinct tasks, and the ability to be integrated into a variety of downstream systems. It does not depend on the sector. The Act has no autonomous systems category and no list of models for it.

  • Provider, system provider and deployer

    The company that develops the model and places it on the market is the provider and carries Article 53. An operator or manufacturer that only uses a tool built on the model is a deployer of an AI system and has no Article 53 duty. A company that integrates the model into its own AI system, for example into a function of a vehicle or robot, and places that system on the market or puts it into service under its own name is the provider of that system; if the system is high-risk, it has the duties of Article 16, and it receives the Annex XII information.

  • Article 2(2) concerns AI systems, not models

    Article 2(2) limits the provisions that apply to AI systems that are high-risk under Article 6(1) and relate to an Annex I Section B product, such as a motor vehicle, aircraft or machine: only Article 6(1), Article 60a and Articles 102 to 112 apply. It speaks of AI systems and does not mention general-purpose AI models. Article 53 binds the provider of the model; on its wording, Article 2(2) does not displace it. Ask counsel whether a Section B product raises any further question.

  • Written agreement with suppliers

    The provider of a high-risk system and a third party that supplies an AI model, tools, services, components or processes used in it must specify by written agreement the information, capabilities, technical access and other assistance needed for full compliance (Article 25(4)). The duty does not apply to third parties that make tools, services, processes or components other than general-purpose AI models publicly accessible under a free and open-source licence.

  • Research exclusions

    The Act does not apply to AI systems or AI models, including their output, specifically developed and put into service for the sole purpose of scientific research and development (Article 2(6)). It does not apply to research, testing or development before a system or model is placed on the market or put into service; testing in real world conditions is not covered by that exclusion (Article 2(8)).

  • Dates

    The duties of Article 53 apply from 2 August 2025 (Article 113, point (b)). A model placed on the market before that date must comply by 2 August 2027 (Article 111(3)). Article 101 is excluded from point (b), so the general date of application, 2 August 2026, governs it.

  • Fines

    The Commission may fine a provider of a general-purpose AI model up to 3% of its total worldwide annual turnover in the preceding financial year or EUR 15 million, whichever is higher, where it finds that the provider intentionally or negligently infringed the relevant provisions, or on another ground listed in Article 101(1). Breaches of the written-agreement duty in Article 25(4) fall under Article 99(4), point (da): up to EUR 15 million or, for an undertaking, 3% of total worldwide annual turnover in the preceding financial year, whichever is higher.

  • Limits of the Act’s scope

    The Act applies to providers that place AI systems on the market or put them into service in the Union, or place general-purpose AI models on the market in the Union, wherever they are established (Article 2(1), point (a)), and to providers and deployers in a third country where the output of the AI system is used in the Union (point (c)). It does not apply to areas outside the scope of Union law and does not affect Member States’ competences on national security, nor to AI systems used exclusively for military, defence or national security purposes (Article 2(3)). Natural persons who use a system in the course of a purely personal non-professional activity are not bound as deployers (Article 2(10)).

What the Act leaves open for autonomous systems

  • No duty specific to autonomous systems

    Article 53 has no duty specific to autonomous systems and names no sector content for the documentation. Annex XI lists the minimum content of the technical documentation (Section 1 for all providers, Section 2 in addition for models with systemic risk) and Annex XII that of the information for downstream providers. The exemption in Article 53(2) can remove the duties in points (a) and (b) for open-source models without systemic risk.

  • Systemic risk is Articles 51, 52 and 55

    The classification of models with systemic risk is in Article 51 and Article 52; their additional duties are in Article 55. Article 53(3) is the duty to cooperate as necessary with the Commission and the national competent authorities.

  • Adapting a model

    The articles of the Act do not say when adapting a model, for example by fine-tuning it on operating data, makes the adapter a provider of a new model; ask counsel. Recitals and Commission guidelines may address the question.

Questions about Article 53 in autonomous systems

Does a vehicle or robot maker that uses a large language model have to meet Article 53?

No. Article 53 binds providers of general-purpose AI models. A maker that integrates a model into its own AI system is the provider of that system, not of the model, and a user of the finished product is a deployer (Article 3(4)).

What must a maker that builds an AI system on a model receive from the model provider?

Information and documentation that let it understand the model’s capabilities and limitations and comply with its own obligations, with at least the elements in Annex XII (Article 53(1), point (b)). Article 25(4) adds a written agreement for a high-risk system.

Is Article 53(3) the systemic-risk rule?

No. Article 53(3) requires providers of general-purpose AI models to cooperate as necessary with the Commission and the national competent authorities. Systemic-risk classification is in Articles 51 and 52, and the additional duties of providers of models with systemic risk are in Article 55.

Do open-source models have to follow Article 53?

In part. Providers of models released under a free and open-source licence with public parameters are exempt from the duties in points (a) and (b), but not from the copyright policy and the training-content summary, and not if the model has systemic risk (Article 53(2)).

Related pages

Scanara prepares evidence from your repositories

Scanara reads your connected GitHub repositories and runs rule-based scans for standard AI and machine-learning patterns. It can miss non-standard code. Separately, it generates draft documentation from templates and your assessment answers as evidence for your compliance work. You make the classification decisions.