Code of Practice for General-Purpose AI Models
Code of Practice for General-Purpose AI Models
The General-Purpose AI (GPAI) Code of Practice is a guiding document for compliance with the AI Act, published on July 10, 2025, as part of a package of documents tied to the entry into application of the AI Act provisions that regulate GPAI models. The Code is prepared on the basis of Article 56 of the AI Act, by independent experts in a multi-stakeholder process, designed to help industry comply with the AI Act’s obligations for providers of GPAI models. On August 1, 2025, the Commission and AI Board approved the code via Adequacy Decisions.
Although not legally binding, GPAI model providers can rely on the Code of Practice to demonstrate compliance with imposed obligations until a harmonised standard is published (AI Act, Article 53(4)). GPAI models providers, who voluntarily sign it, can show they comply with the AI Act by adhering to the Code. Signing the Code is done by completing the Signatory Form and sending it to the indicated AI Office internet address.[1] However, accepting to adhere to the Code does not constitute conclusive evidence of compliance with the obligations under the AI Act.[2]
The specific objective of this Code is also to enable the AI Office to assess compliance of providers of GPAI models who choose to rely on the Code. Providers of GPAI models, who do not adhere to the Code of practice, have to prove compliance with their obligations to the Commission by alternative adequate, possibly more burdensome and time-consuming means.
The Code consists of three chapters: Transparency, Copyright, and Safety and Security.[3] The first two, Transparency and Copyright, apply to all GPAI model providers. The third, Safety and Security chapter, only applies to providers of GPAI models with systemic risk. The Code determines a total of 12 commitments – one in each of the first two chapters and 10 in the Safety and Security Chapter, as well as corresponding measures for fulfilling these commitments.
[1] https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai.
[2] Code of Practice for GPAI Models, the Transparency Chapter, Objectives, page 3.
[3] https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai.
Transparency Chapter
The Transparency chapter details the documentation obligations from Article 53(1), points (a) and (b) of the EU AI Act, which require providers of GPAI models to:
- 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;
- draw up, keep up-to-date and make available information and documentation to providers of AI systems who intend to integrate the GPAI 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:
- enable providers of AI systems to have a good understanding of the capabilities and limitations of the GPAI model and to comply with their obligations pursuant to this Regulation; and
- contain, at a minimum, the elements set out in Annex XII.
Exeption: these obligations will not apply to providers of GPAI 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 will not apply to GPAI models with systemic risks.
Commitment 1 Documentation
(Articles 53(1)(a), 53(1)(b), 53(2), 53(7), and Annexes XI and XII AI Act)
The Transparency chapter includes three measures for Signatories to commit to:
Measure 1.1 Drawing up and keeping up-to-date model documentation
Signatories, when placing a GPAI model on the market, will have documented at least all the information referred to in a standardized Model Documentation Form, detailing licensing, technical specs, use cases, datasets, compute and energy usage, and more.
Measure 1.2 Providing relevant information
Signatories, when placing a GPAI on the market, will publicly disclose via their website, or via other appropriate means if they do not have a website, contact information for the AI Office and downstream providers to request access to the relevant information contained in the Model Documentation, or other necessary information.
Measure 1.3 Ensuring quality, integrity, and security of information
Signatories will ensure that the documented information is controlled for quality and integrity, retained as evidence of compliance with obligations in the AI Act, and protected from unintended alterations.
Copyright Chapter
This Chapter aims to contribute to the proper application of the provision in Article 53(1), point (c), of the AI Act pursuant to which providers that place GPAI models on the Union market must 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 by rightsholders pursuant to Article 4(3) of Directive (EU) 2019/790.
GPAI, capable of generating text, images, and other content, presents unique innovation opportunities but also challenges to artists, authors, and other creators and the way their creative content is created, distributed, used and consumed. The development and training of such models require access to vast amounts of text, images, videos, and other data. Text and data mining techniques may be used extensively in this context for the retrieval and analysis of such content, which may be protected by copyright and related rights. Any use of copyright-protected content requires the authorization of the rightholder concerned unless relevant copyright exceptions and limitations apply. Directive (EU) 2019/790 introduced exceptions and limitations allowing reproductions and extractions of works or other subject matter, for the purposes of text and data mining, under certain conditions. Under these rules, rightholders may choose to reserve their rights over their works or other subject matter to prevent text and data mining, unless this is done for the purposes of scientific research. Where the rights to opt out has been expressly reserved in an appropriate manner, providers of GPAI models need to obtain an authorisation from rightholders if they want to carry out text and data mining over such works.
In this chapter, it is explicitly stated that the envisaged measures in no way affect the application and enforcement of Union law on copyright and related rights, which is for the courts of Member States and ultimately the Court of Justice of the European Union to interpret.
The Signatories hereby acknowledge that Union law on copyright and related rights:
- is provided for in directives addressed to Member States and that, for the purposes of this Code, Directive 2001/29/EC[1], Directive (EU) 2019/790[2] and Directive 2004/48/EC[3] are the most relevant;
- provides for exclusive rights that are preventive in nature and thus is based on prior consent, save where an exception or limitation applies;
- provides for an exception or limitation for text and data mining in Article 4(1) of Directive (EU) 2019/790 which shall apply on conditions of lawful access and that the use of works and other subject matter referred to in that paragraph has not been expressly reserved by their rightsholders in an appropriate manner pursuant to Article 4(3) of Directive (EU) 2019/790.[4]
Commitment 1 Copyright policy
(Article 53(1)(c) AI Act)
Signatories commit to drawing up, keeping up-to-date and implementing a copyright policy. Envisaged measures do not affect compliance with Union law on copyright and related rights. They set out commitments by which the Signatories can demonstrate compliance with the obligation to put in place a copyright policy for their GPAI models they place on the Union market.
In addition, the Signatories remain responsible for verifying that the Measures, included in their copyright policy, comply with Member States’ implementation of Union law on copyright and related rights, in particular Article 4(3) of Directive (EU) 2019/790, before carrying out any copyright-relevant act in the territory of the relevant Member State, as failure to do so may give rise to liability under Union law on copyright and related rights.
Measure 1.1 Draw up, keep up-to-date and implement a copyright policy
(1) Signatories will draw up, keep up-to-date and implement a policy to comply with Union law on copyright and related rights for all GPAI models they place on the Union market. Signatories commit to describing that policy in a single document incorporating the Measures set out in this Chapter, and assign responsibilities within their organisation for the implementation and overseeing of this policy.
(2) Signatories are encouraged to make publicly available and keep up-to-date a summary of their copyright policy.
Measure 1.2 Reproduce and extract only lawfully accessible copyright-protected content when crawling the World Wide Web
(1) To help ensure that Signatories only reproduce and extract lawfully accessible works and other protected content if they use web-crawlers or have such web-crawlers used on their behalf to scrape or otherwise compile data for the purpose of text and data mining, as defined in Article 2(2) of Directive (EU) 2019/790, and the training of their general-purpose AI models, Signatories commit:
a) not to circumvent effective technological measures as defined in Article 6(3) of Directive 2001/29/EC that are designed to prevent or restrict unauthorised acts in respect of works and other protected, in particular by respecting any technological denial or restriction of access imposed by subscription models or paywalls, and
b) to exclude from their web-crawling websites that make available to the public content and which are, at the time of web-crawling, recognised as persistently and repeatedly infringing copyright and related rights on a commercial scale.
Measure 1.3 Identify and comply with rights reservations when crawling the World Wide Web
In order to help ensure that Signatories will identify and comply with, including through state-of-the-art technologies, machine-readable reservations of rights expressed pursuant to Article 4(3) of Directive (EU) 2019/790 if they use web-crawlers or have such web-crawlers used on their behalf to scrape or otherwise compile data for the purpose of text and data mining as defined in Article 2(2) of Directive (EU) 2019/790 and the training of their GPAI models, Signatories commit:
a) to employ web-crawlers that read and follow instructions expressed in accordance with the Robot Exclusion Protocol (robots.txt), as specified in the Internet Engineering Task Force (IETF) Request for Comments No. 9309, and any subsequent version of this Protocol for which the IETF demonstrates that it is technically feasible and implementable by AI providers and content providers, including rightsholders, and
b) to identify and comply with other appropriate machine-readable protocols to express rights reservations pursuant to Article 4(3) of Directive (EU) 2019/790, for example through asset-based or location-based metadata, that have either have been adopted by international or European standardisation organisations, or are state-of-the-art, including technically implementable, and widely adopted by rightsholders, considering different cultural sectors, and generally agreed through an inclusive process based on bona fide discussions to be facilitated at EU level with the involvement of rightsholders, AI providers and other relevant stakeholders as a more immediate solution, while anticipating the development of standards.
Measure 1.4 Mitigate the risk of copyright-infringing outputs
In order to mitigate the risk that a downstream AI system, into which a general-purpose AI model is integrated, generates output that may infringe rights in works or other subject matter protected by Union law on copyright or related rights, Signatories commit:
a) to implement appropriate and proportionate technical safeguards to prevent their models from generating outputs that reproduce training content protected by Union law on copyright and related rights in an infringing manner, and
b) to prohibit copyright-infringing uses of a model in their acceptable use policy, terms and conditions, or other equivalent documents, or in case of general-purpose AI models released under free and open source licenses to alert users to the prohibition of copyright infringing uses of the model in the documentation accompanying the model without prejudice to the free and open source nature of the license.
Measure 1.5 Designate a point of contact and enable the lodging of complaints
Signatories commit to designating a point of contact for electronic communication with affected rightsholders and providing easily accessible information about it.
Signatories commit to put a mechanism in place to allow affected rightsholders and their authorised representatives, including collective management organisations, to submit, by electronic means, sufficiently precise and adequately substantiated complaints concerning the non-compliance of Signatories with their commitments pursuant to this Chapter and provide easily accessible information about it. Signatories will act on complaints in a diligent, non-arbitrary manner and within a reasonable time, unless a complaint is manifestly unfounded or the Signatory has already responded to an identical complaint by the same rightsholder. This commitment does not affect the measures, remedies and sanctions available to enforce copyright and related rights under Union and national law.
[1] Directive 2001/29/EC of the European Parliament and of the Council of 22 May 2001 on the harmonisation of certain aspects of copyright and related rights in the information society.
[2] Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on copyright and related rights in the Digital Single Market and amending Directives 96/9/EC and 2001/29/EC.
[3] Directive 2004/48/EC of the European Parliament and of the Council of 29 April 2004 on the enforcement of intellectual property rights.
[4] The Code, Copyright Chapter, page 3.
Safety and Security Chapter
The Safety and Security chapter of the Code is relevant to entities providing GPAI models with systemic risk. ‘Systemic risk’ is defined in AI Act as specific to the high-impact capabilities of GPAI 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 3(65)).
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:
- it has high impact capabilities evaluated on the basis of appropriate technical tools and methodologies, including indicators and benchmarks;
- 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 51).
Models are presumed to qualify as having high impact capabilities when the cumulative amount of computation used for its training is greater than 10^25 floating-point operations (‘FLOP’)[1]. This is a rebuttable presumption. Currently, over 30 publicly announced AI models from different AI developers have been identified, which surpass this threshold.[2]
In addition to the obligations of providers of other GPAI, pursuant to provisions of Article 55 of the AI Act, providers of GPAI models with systemic risk must:
(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 the 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.
The Safety and Security Chapter determines 10 commitments and corresponding measures for fulfilling these commitments. In the following, these commitments and corresponding measures, due to the length of the text, are only listed, while the full text is available on the Commission’s website[3].
Commitment 1 Safety and Security Framework
(Articles 55(1) and 56(5), and recitals 110, 114, and 115 AI Act)
Signatories commit to adopting a state-of-the-art Safety and Security Framework (“Framework”). The purpose of the Framework is to outline the systemic risk management processes and measures that Signatories implement to ensure the systemic risks stemming from their models are acceptable.
Measure 1.1 Creating the Framework
The Framework will contain a high-level description of implemented and planned processes and measures for systemic risk assessment and mitigation to adhere to this Chapter.
Measure 1.2 Implementing the Framework
Measure 1.3 Updating the Framework
Measure 1.4 Framework notifications
Signatories will provide the AI Office with (unredacted) access to their Framework, and updates thereof, within five business days of either being confirmed.
Commitment 2 Systemic risk identification
(Article 55(1) and recital 110 AI Act)
Signatories commit to identifying the systemic risks stemming from the model, for the purpose of facilitating systemic risk analysis (pursuant to Commitment 3) and
systemic risk acceptance determination (pursuant to Commitment 4).
Measure 2.1 Systemic risk identification process
Measure 2.2 Systemic risk scenarios
Signatories will develop appropriate systemic risk scenarios, including regarding the number and level of detail of these systemic risk scenarios, for each identified systemic risk (pursuant to Measure 2.1).
Commitment 3 Systemic risk analysis
(Article 55(1) and recital 114 AI Act)
Signatories commit to analysing each identified systemic risk (pursuant to Commitment 2) for the purpose of facilitating systemic risk acceptance determination (pursuant to Commitment 4).
Measure 3.1 Model-independent information
Measure 3.2 Model evaluations
Measure 3.3 Systemic risk modelling
Measure 3.4 Systemic risk estimation
Measure 3.5 Post-market monitoring
Commitment 4 Systemic risk acceptance determination
(Article 55(1) AI Act)
Signatories commit to specifying systemic risk acceptance criteria and determining whether the systemic risks stemming from the model are acceptable (as specified in Measure 4.1). Signatories commit to deciding whether or not to proceed with the development, the making available on the market, and/or the use of the model based on the systemic risk acceptance determination (as specified in Measure 4.2).
Measure 4.1 Systemic risk acceptance criteria and acceptance determination
Measure 4.2 Proceeding or not proceeding based on systemic risk acceptance determination
Commitment 5 Safety mitigations
(Article 55(1) and recital 114 AI Act)
Signatories commit to implementing appropriate safety mitigations along the entire model lifecycle, as specified in the Measure for this Commitment, to ensure the systemic risks stemming from the model are acceptable (pursuant to Commitment 4).
Measure 5.1 Appropriate safety mitigations
Commitment 6 Security mitigations
(Article 55(1), and recitals 114 and 115 AI Act)
Measure 6.1 Security Goal
Measure 6.2 Appropriate security mitigations
Commitment 7 Safety and Security Model Reports
(Articles 55(1) and 56(5) AI Act)
Signatories commit to reporting to the AI Office information about their model and their systemic risk assessment and mitigation processes and measures by creating a Safety and Security Model Report (“Model Report”) before placing a model on the market, to keep the Model Report up-to-date and to notify the AI Office of their Model Report.
Measure 7.1 Model description and behaviour
Measure 7.2 Reasons for proceeding
Measure 7.3 Documentation of systemic risk identification, analysis, and mitigation
Measure 7.4 External reports
Signatories will provide in the Model Report:
(1) any available reports (e.g. via valid hyperlinks) from:
(a) independent external evaluators involved in model evaluations pursuant to Appendix3.5; and
(b) security reviews undertaken by an independent external party pursuant to Appendix 4.5.
Measure 7.5 Material changes to the systemic risk landscape
Measure 7.6 Model Report updates
Measure 7.7 Model Report notifications
Commitment 8 Systemic risk responsibility allocation
(Article 55(1) and recital 114 AI Act)
Signatories commit to: (1) defining clear responsibilities for managing the systemic risks stemming from their models across all levels of the organisation (as specified in Measure 8.1); (2) allocating appropriate resources to actors who have been assigned responsibilities for managing systemic risk (as specified in Measure 8.2); and (3) promoting a healthy risk culture (as specified in Measure 8.3).
Measure 8.1 Definition of clear responsibilities
Measure 8.2 Allocation of appropriate resources
Measure 8.3 Promotion of a healthy risk culture
Commitment 9 Serious incident reporting
(Article 55(1), and recitals 114 and 115 AI Act)
Measure 9.1 Methods for serious incident identification
Measure 9.2 Relevant information for serious incident tracking, documentation, and Reporting
Measure 9.3 Reporting timelines
Measure 9.4 Retention period
Commitment 10 Additional documentation and transparency
(Articles 53(1)(a) and 55(1) AI Act)
Measure 10.1 Additional documentation
Measure 10.2 Public transparency