General-purpose AI models

Chapter V of the EU AI Act (Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744) gives obligations to providers of general-purpose AI (GPAI) models. Chapter V has these parts:

• Articles 51 and 52 — the classification of GPAI models with systemic risk, and its procedure.
• Article 53 — the obligations of all providers of GPAI models.
• Article 54 — authorized representatives of providers in third countries.
• Article 55 — more obligations for providers of GPAI models with systemic risk.
• Article 56 — codes of practice.

Scope and dates

The EU AI Act applies to providers "placing on the market general-purpose AI models in the Union, irrespective of whether those providers are established or located within the Union or in a third country" (Article 2(1), point (a)). Article 2 has different tests for other actors, and it has exclusions and limits. See Article 2 — scope.

Chapter V applies to GPAI model providers from 2 August 2025, with transitional rules for earlier models and different enforcement dates (Articles 111(3) and 113). See Dates.

Article 3(63) — general-purpose AI model

For GPAI model providers, Article 3(63) of the EU AI Act gives this definition:

‘general-purpose AI model’ means an AI model, including where such an AI model is trained with a large amount of data using self-supervision at scale, that displays significant generality and is capable of competently performing a wide range of distinct tasks regardless of the way the model is placed on the market and that can be integrated into a variety of downstream systems or applications, except AI models that are used for research, development or prototyping activities before they are placed on the market;

Thus an AI model is a GPAI model only if all of these conditions apply:

• The model "displays significant generality".
• The model "is capable of competently performing a wide range of distinct tasks". This test applies "regardless of the way the model is placed on the market".
• The model "can be integrated into a variety of downstream systems or applications".

The definition excludes AI models "that are used for research, development or prototyping activities before they are placed on the market".

The Commission gives an indicative criterion for this test in its GPAI-scope Guidelines, Communication "C(2025) 7719 final" of 19 November 2025 (paragraph (17)). On this criterion, a model is considered a GPAI model when two conditions apply. Its training compute is more than 10²³ floating-point operations. It can generate language (as text or audio), or it can do text-to-image or text-to-video generation. Paragraph (20) gives exceptions in the two directions:

• If a model "meets the criterion" but, "exceptionally, does not display significant generality or is not capable of competently performing a wide range of distinct tasks", it is not a GPAI model.
• If a model "does not meet that criterion" but, "exceptionally, displays significant generality and is capable of competently performing a wide range of distinct tasks", it is a GPAI model.

The Guidelines are not binding (paragraph (9)). The 10²³ criterion is not the 10²⁵ presumption for systemic risk in Article 51(2). For the Commission interpretation, see The GPAI model test.

A GPAI model is not an AI system. Article 3(66) gives a different definition for a general-purpose AI system: "an AI system which is based on a general-purpose AI model and which has the capability to serve a variety of purposes, both for direct use as well as for integration in other AI systems". For the definition of an AI system, see Article 3 — definitions.

The EU AI Act gives different rules for models and for systems:

• GPAI models — Chapter V gives obligations to providers of GPAI models placed on the market in the Union (Articles 2(1)(a) and 53). The classification of a model in Article 51 adds the obligations in Article 55.
• AI systems — Articles 5, 6 and 50 apply to AI systems, also to general-purpose AI systems. Article 6 gives the conditions that make an AI system high-risk. The use of a GPAI model is not one of these conditions. See Article 6 — classification rules for high-risk AI systems.

The integration of a GPAI model into an AI system does not, by itself, make the system provider subject to Chapter V. That provider has the obligations for its AI system. A downstream actor can also change a GPAI model. The EU AI Act does not give the conditions in which this actor becomes the provider (paragraph (57) of the GPAI-scope Guidelines). In the Commission interpretation, this occurs "only if the modification leads to a significant change in the model's generality, capabilities, or systemic risk" (paragraph (59)). See When a model modifier becomes a provider.

One entity can have the obligations for a model and the obligations for a system. Article 3(68) defines a downstream provider as a provider of an AI system "which integrates an AI model, regardless of whether the AI model is provided by themselves and vertically integrated or provided by another entity based on contractual relations". Article 75(1), point (a), gives the supervision rules for AI systems where the same provider develops the model and the system (see Enforcement).

An AI system that integrates a GPAI model can be high-risk. Then its provider must comply with Article 16 and make sure that the system complies with Articles 8 to 15. Article 2(2) excludes these provisions for high-risk AI systems in Article 6(1) that are related to products in Section B of Annex I. The Annex XII information from the model provider helps the system provider, but it does not make the system compliant.

Article 25 of the EU AI Act gives the conditions for changes in provider responsibility for high-risk AI systems.

Article 51 — classification of GPAI models with systemic risk

For GPAI model providers, Article 51(1) of the EU AI Act gives two conditions for classification with systemic risk. One condition is sufficient:

A general-purpose AI model shall be classified as a general-purpose AI model with systemic risk if it meets any of the following conditions:

(a) it has high impact capabilities evaluated on the basis of appropriate technical tools and methodologies, including indicators and benchmarks;

(b) based on a decision of the Commission, ex officio or following a qualified alert from the scientific panel, it has capabilities or an impact equivalent to those set out in point (a) having regard to the criteria set out in Annex XIII.

Article 3(64) defines high-impact capabilities as "capabilities that match or exceed the capabilities recorded in the most advanced general-purpose AI models". Article 3(65) gives this definition of systemic risk:

‘systemic risk’ means a risk that is specific to the high-impact capabilities of general-purpose AI models, having a significant impact on the Union market due to their reach, or due to actual or reasonably foreseeable negative effects on public health, safety, public security, fundamental rights, or the society as a whole, that can be propagated at scale across the value chain;

Article 51(2) — the 10²⁵ FLOPs presumption

Article 51(2) of the EU AI Act gives a presumption for the condition in Article 51(1), point (a). A GPAI model is presumed to have high-impact capabilities "when the cumulative amount of computation used for its training measured in floating point operations is greater than" 10²⁵.

The training compute gives only a presumption of high-impact capabilities. A provider can give arguments against the classification, and the Commission decides on them (Article 52(2) and (3)). A model with a lower training compute can also have systemic risk. This occurs when it has high-impact capabilities (Article 51(1), point (a)), or when the Commission designates it (Article 51(1), point (b)). The Commission can amend the thresholds, and supplement the benchmarks and indicators, with delegated acts (Article 51(3)).

Article 52 — procedure for designation

Article 52 of the EU AI Act gives GPAI model providers and the Commission this procedure for classification with systemic risk:

• Notification (Article 52(1)) — When the condition in Article 51(1), point (a), applies to a GPAI model, the provider must notify the Commission. The provider must do this "without delay and in any event within two weeks after that requirement is met or it becomes known that it will be met". The notification must include "the information necessary to demonstrate that the relevant requirement has been met". If the Commission learns of a GPAI model "presenting systemic risks of which it has not been notified", it "may decide to designate it as a model with systemic risk".
• Arguments against the classification (Article 52(2)) — With its notification, the provider can give "sufficiently substantiated arguments to demonstrate that, exceptionally, although it meets that requirement, the general-purpose AI model does not present, due to its specific characteristics, systemic risks".
• Arguments rejected (Article 52(3)) — If the Commission concludes that these arguments are not sufficiently substantiated, it rejects them. The model is then a GPAI model with systemic risk.
• Designation (Article 52(4)) — The Commission can designate a GPAI model "as presenting systemic risks". It does this ex officio or after a qualified alert from the scientific panel (Article 90(1), point (a)), with the criteria in Annex XIII. The Commission can amend Annex XIII with delegated acts.
• Publication (Article 52(6)) — The Commission "shall ensure that a list of general-purpose AI models with systemic risk is published and shall keep that list up to date". This obligation is without prejudice to the protection of intellectual property rights, confidential business information and trade secrets.

Article 52(5) gives a reassessment procedure only to the provider of a model that the Commission designated in Article 52(4). These rules apply:

• The provider sends a "reasoned request". The reasoned request must contain "objective, detailed and new reasons that have arisen since the designation decision".
• The provider can send the reasoned request no earlier than six months after the designation decision. If the Commission reassesses the model and keeps the designation, the provider can send a new reasoned request no earlier than six months after that decision.
• The Commission "shall take the request into account". The Commission decides if it reassesses the model, on the criteria in Annex XIII.

For the Commission interpretation of notification and arguments, see Classification of systemic risk.

Article 53 — obligations for all GPAI model providers

Article 53(1) of the EU AI Act gives four obligations to all providers of GPAI models, with or without systemic risk:

Providers of general-purpose AI models shall:

(a) draw up and keep up-to-date the technical documentation of the model, including its training and testing process and the results of its evaluation, which shall contain, at a minimum, the information set out in Annex XI for the purpose of providing it, upon request, to the AI Office and the national competent authorities;

(b) draw up, keep up-to-date and make available information and documentation to providers of AI systems who intend to integrate the general-purpose AI model into their AI systems. Without prejudice to the need to observe and protect intellectual property rights and confidential business information or trade secrets in accordance with Union and national law, the information and documentation shall:

(i) enable providers of AI systems to have a good understanding of the capabilities and limitations of the general-purpose AI model and to comply with their obligations pursuant to this Regulation; and

(ii) contain, at a minimum, the elements set out in Annex XII;

(c) put in place a policy to comply with Union law on copyright and related rights, and in particular to identify and comply with, including through state-of-the-art technologies, a reservation of rights expressed pursuant to Article 4(3) of Directive (EU) 2019/790;

(d) draw up and make publicly available a sufficiently detailed summary about the content used for training of the general-purpose AI model, according to a template provided by the AI Office.

Each obligation has a different recipient:

Article 53(1)	Recipient	Minimum content
Point (a), technical documentation	The AI Office and the national competent authorities, "upon request"	Annex XI
Point (b), information and documentation for integration	Providers of AI systems that intend to integrate the model	Annex XII
Point (c), copyright policy	No recipient in Article 53(1)	Union law on copyright and related rights, and Article 4(3) of Directive (EU) 2019/790
Point (d), summary of training content	The public	The template of the AI Office

The Commission gives the template for point (d), with an Explanatory Notice, in Communication "C(2025) 8311 final" of 5 December 2025.

Article 53 of the EU AI Act also has these rules:

• Providers of GPAI models must "cooperate as necessary with the Commission and the national competent authorities" (Article 53(3)).
• Providers can use codes of practice, harmonized standards or "alternative adequate means of compliance" (Article 53(4)). See Article 56 — Codes of Practice.
• The Commission can adopt delegated acts on "measurement and calculation methodologies" for Annex XI (Article 53(5)). It can also amend Annexes XI and XII with delegated acts (Article 53(6)).
• Information and documentation that the authorities get in Article 53, also trade secrets, is subject to the confidentiality obligations of Article 78 (Article 53(7)).

Article 53(2) — free-and-open-source exception

Article 53(2) of the EU AI Act gives an exception for some open-source models:

The obligations set out in paragraph 1, points (a) and (b), shall not apply to providers of AI models that are released under a free and open-source licence that allows for the access, usage, modification, and distribution of the model, and whose parameters, including the weights, the information on the model architecture, and the information on model usage, are made publicly available. This exception shall not apply to general-purpose AI models with systemic risks.

The exception applies only if all of these conditions apply:

• The license "allows for the access, usage, modification, and distribution of the model".
• The parameters of the model, "including the weights, the information on the model architecture, and the information on model usage, are made publicly available".
• The model is not a GPAI model with systemic risk.

The exception covers only points (a) and (b) of Article 53(1). The copyright policy (point (c)) and the summary of training content (point (d)) apply to all providers of GPAI models. Article 54(6) gives an exception with the same conditions for the authorized representative (see Article 54). Article 2(12) gives a different exclusion for "AI systems released under free and open-source licences". For the Commission interpretation of the license conditions, see Free and open-source models.

Article 54 — authorized representatives for non-EU GPAI providers

Article 54(1) of the EU AI Act gives an obligation to providers of GPAI models that are established in third countries. They must appoint an authorized representative established in the Union, by written mandate. They must do this "Prior to placing a general-purpose AI model on the Union market". The provider must let its representative do the tasks in the mandate (Article 54(2)).

The authorized representative must do the tasks specified in the mandate received from the provider (Article 54(3)). The mandate must let the representative do these tasks:

• (a) "verify that the technical documentation specified in Annex XI has been drawn up and all obligations referred to in Article 53 and, where applicable, Article 55 have been fulfilled by the provider".
• (b) Keep a copy of the technical documentation in Annex XI for the AI Office and national competent authorities. Keep it for 10 years after the placing on the market of the model. Also keep the contact details of the provider that appointed the representative.
• (c) On a reasoned request, give the AI Office all the information and documentation necessary to show compliance with Chapter V. This includes the documentation in point (b).
• (d) On a reasoned request, cooperate with the AI Office and competent authorities "in any action they take in relation to the general-purpose AI model". This includes the integration of the model into AI systems placed on the market or put into service in the Union.

The authorized representative of a GPAI model provider also has these rules:

• The representative must give a copy of the mandate to the AI Office "upon request". The copy must be in one of the official languages of the institutions of the Union (Article 54(3)).
• The mandate lets the AI Office or the competent authorities address the representative on all issues of compliance with the EU AI Act. They can address it "in addition to or instead of the provider" (Article 54(4)).
• The representative must terminate the mandate "if it considers or has reason to consider the provider to be acting contrary to its obligations pursuant to this Regulation". It must then immediately inform the AI Office about the termination and the reasons (Article 54(5)).

Article 54 does not apply to providers of GPAI models released "under a free and open-source licence", on the conditions of Article 53(2). This exception does not apply when the models "present systemic risks" (Article 54(6)).

Article 22 of the EU AI Act gives different rules for authorized representatives of providers of high-risk AI systems. A provider in a third country can be the provider of a high-risk AI system and of a GPAI model. Then Article 22 and Article 54 can apply together, each on its own conditions. See Article 22 — authorized representatives for non-EU providers of high-risk AI systems.

Article 55 — obligations for systemic-risk GPAI

Article 55(1) of the EU AI Act gives more obligations to providers of GPAI models with systemic risk:

In addition to the obligations listed in Articles 53 and 54, providers of general-purpose AI models with systemic risk shall:

(a) perform model evaluation in accordance with standardised protocols and tools reflecting the state of the art, including conducting and documenting adversarial testing of the model with a view to identifying and mitigating systemic risks;

(b) assess and mitigate possible systemic risks at Union level, including their sources, that may stem from the development, the placing on the market, or the use of general-purpose AI models with systemic risk;

(c) keep track of, document, and report, without undue delay, to the AI Office and, as appropriate, to national competent authorities, relevant information about serious incidents and possible corrective measures to address them;

(d) ensure an adequate level of cybersecurity protection for the general-purpose AI model with systemic risk and the physical infrastructure of the model.

Point (c) gives the serious-incident obligation of the model provider. The provider reports to the AI Office and, "as appropriate", to national competent authorities. A national competent authority is a notifying authority or a market surveillance authority (Article 3(48)). The time limit is "without undue delay", and Article 55(1) gives no number of days. Article 3(49) defines a serious incident. See Article 3(49) — serious incident.

A model provider's report in Article 55(1)(c) does not replace an applicable Article 73 report by a high-risk AI system provider.

Providers of GPAI models with systemic risk can use codes of practice, harmonized standards or "alternative adequate means of compliance" for Article 55(1) (Article 55(2)). See Article 56 — Codes of Practice. Information and documentation that the authorities get in Article 55, also trade secrets, is subject to the confidentiality obligations of Article 78 (Article 55(3)).

Article 56 — Codes of Practice

Article 56 of the EU AI Act gives the AI Office rules for codes of practice for GPAI model providers. Adherence to a code of practice is voluntary (Article 56(7)). A code helps providers to demonstrate compliance with the obligations in Articles 53 and 55, but it does not replace these obligations.

Articles 53(4) and 55(2) of the EU AI Act give three ways to show compliance with Articles 53(1) and 55(1):

• 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". A code of practice does not give a presumption of conformity (recital 41 of Regulation (EU) 2026/1744).
• Harmonized standards — "Compliance with European harmonised standards grants providers the presumption of conformity to the extent that those standards cover those obligations."
• Other means — Providers "who do not adhere to an approved code of practice or do not comply with a European harmonised standard shall demonstrate alternative adequate means of compliance for assessment by the Commission".

Conformity with harmonized standards and common specifications gives a presumption only for obligations in their scope, on the conditions in Articles 40(1) and 41(3).

Article 56 of the EU AI Act gives this procedure for codes of practice:

• The AI Office encourages and facilitates the drawing up of codes of practice at Union level (Article 56(1)). It takes international approaches into account.
• The AI Office and the AI Board "shall aim to ensure that the codes of practice cover at least the obligations provided for in Articles 53 and 55". Article 56(2) gives four issues for the codes, for example "the adequate level of detail for the summary about the content used for training".
• The AI Office can invite "all providers of general-purpose AI models, as well as relevant national competent authorities" to participate. Article 56(3) adds: "Civil society organisations, industry, academia and other relevant stakeholders, such as downstream providers and independent experts, may support the process."
• The AI Office and the AI Board "shall aim to ensure" that the codes set out their objectives and contain commitments or measures, "including key performance indicators as appropriate" (Article 56(4)). The AI Office "shall aim to ensure" that participants report regularly to it (Article 56(5)).
• The Commission and the AI Board "shall regularly monitor and evaluate the achievement of the objectives of the codes of practice". The Commission, "taking utmost account of the opinion of the Board", "shall assess whether the codes of practice cover the obligations provided for in Articles 53 and 55". It must publish "its assessment of the adequacy of the codes of practice" (Article 56(6), as replaced by Regulation (EU) 2026/1744).
• The AI Office "may invite all providers of general-purpose AI models to adhere to the codes of practice". Article 56(7) adds: "For providers of general-purpose AI models not presenting systemic risks this adherence may be limited to the obligations provided for in Article 53, unless they declare explicitly their interest to join the full code."
• The AI Office "shall, as appropriate, also encourage and facilitate the review and adaptation of the codes of practice, in particular in light of emerging standards" (Article 56(8)).

Article 56(9) tells the AI Office that codes of practice "shall be ready at the latest by 2 May 2025". The Commission can give "common rules for the implementation of the obligations provided for in Articles 53 and 55" by implementing acts, when one of these conditions applies:

• "by 2 August 2025, a code of practice cannot be finalised"
• the AI Office "deems it is not adequate following its assessment under paragraph 6". The date of 2 August 2025 does not apply to this second condition.

The Commission page "The General-Purpose AI Code of Practice" (last update: 7 October 2026) gives 10 July 2025 as the publication date of the GPAI Code of Practice. The Commission page of 1 August 2025 on its opinion tells providers that the Commission and the AI Board confirmed that the Code is "an adequate voluntary tool for providers of GPAI models to demonstrate compliance with the AI Act". The Transparency and Copyright chapters of the Code are for the obligations in Article 53. Its Safety and Security chapter is for the obligations in Article 55. For the Commission interpretation of codes, see Codes of Practice.

Enforcement

The Commission has exclusive powers to supervise and enforce Chapter V of the EU AI Act through the AI Office (Article 88(1)). See Penalties and enforcement for powers, fines, complaints and system-supervision exceptions.

Modulos

EU AI Act GPAI Requirements in Modulos are in OFF-1 at Organization level and MFF-1 at AI System level (see setup steps and the Requirement table).

Related information

• Frequently asked questions about the EU AI Act
• EU AI Act overview
• Roles and responsibilities
• High-risk AI systems
• Post-market monitoring and serious incidents
• Commission guidance on general-purpose AI models
• Operationalizing the EU AI Act in Modulos
• Compare frameworks

Sources

• EU AI Act: Regulation (EU) 2024/1689 of 13 June 2024, consolidated version of 27 July 2026. EUR-Lex.
  • Articles 2(1)–(2), 2(12), 3(48)–(49), 3(63)–(66), and 3(68).
  • Articles 6, 16, 22, 25(1), 40(1), and 41(3).
  • Articles 51 to 56, 73(1) and (10), 75(1) and (1a), and 78.
  • Articles 88(1), 89(2), 91 to 93, 101(1), 111(3), and 113.
  • Annexes XI, XII, and XIII.
• Digital Omnibus on AI: Regulation (EU) 2026/1744 of 8 July 2026, published in the Official Journal on 24 July 2026. Article 1, point (21), which replaces Article 56(6), and recital 41. EUR-Lex.
• GPAI-scope Guidelines: "Commission Guidelines on the scope of the obligations for providers of general-purpose AI models established by Regulation (EU) 2024/1689 (AI Act)", Communication from the Commission "C(2025) 7719 final", 19 November 2025. Paragraphs (9), (17), (20), (57), and (59). Commission text.
• Training summary: "Explanatory Notice and Template for the Public Summary of Training Content for general-purpose AI models required by Article 53 (1)(d) of Regulation (EU) 2024/1689 (AI Act)", Communication from the Commission "C(2025) 8311 final", 5 December 2025. Commission text.
• General-Purpose AI Code of Practice: Commission page "The General-Purpose AI Code of Practice", last update 7 October 2026. Commission page. Commission page "Commission Opinion on the assessment of the General-Purpose AI Code of Practice", 1 August 2025. Commission page.

Disclaimer

This page gives general information. It is not legal advice.