Views: 261
The code of practice represents a voluntary tool prepared by independent experts designed to help various users and industrial community to comply with the AI Act’s rules on general-purpose AI. The AI Act rules on general-purpose AI apply from 2 August 2025; the AI Office has been facilitating the drawing-up of a Code of Practice to detail out these rules.
Background
The code of practice helps industry comply with the AI Act legal obligations on safety, transparency and copyright of general-purpose AI models. So far, the General-Purpose AI (GPAI) code of practice, which is a voluntary tool, prepared by independent experts in a multi-stakeholder process, has been designed to help industry comply with the AI Act’s obligations for providers of general-purpose AI models.
The GPAI code of practice (received and approved by the Commission) details the AI Act rules for providers of general-purpose AI models and general-purpose AI models with systemic risks.
Besides, the European AI Office facilitated the drawing-up of the Code, chaired by independent experts, involving nearly 1000 stakeholders, as well as EU member states representatives, European and international observers.
The code was published in July 2025; then, lately, the EU states and the Commission will be assessing the code and law’s adequacy; additionally, the code will be complemented by Commission guidelines on key concepts related to general-purpose AI models published this July.
More information on the code in: https://digital-strategy.ec.europa.eu/en/policies/ai-code-practice
The code’s narrative
The Code has been prepared in an iterative drafting process: thus, in September2024, the AI Office hosted the inaugural event for the Code of Practice Plenary involving nearly 1000 participants, including a high number of professional organizations. The drafting followed the eligible respondents to the call to participate in the drawing-up launched by the AI Office in July 2024.
The plenary was structured in four working groups on specific topics; and the participants were convened three times virtually for drafting rounds with discussions organized in working groups. For each of the working groups, discussions were facilitated by chairs and vice-chairs, who were selected and appointed by the AI Office from interested independent experts. Participants were able to express comments during each of those meetings or within two weeks in writing.
The Code has been drafted in an inclusive and transparent process, with a variety of interested stakeholders involved: general-purpose AI model providers, downstream providers, industry organizations, civil society, rightsholders and other entities, as well as academia and independent experts.
More in: https://digital-strategy.ec.europa.eu/en/news/kick-plenary-general-purpose-ai-code-practice-took-place-online
As the main addressees of the Code, i.e. the providers of general-purpose AI models were invited to dedicated workshops with the Chairs and Vice-Chairs, contributing to informing each iterative drafting round, in addition to their Plenary participation. The AI Office has ensured transparency into these discussions.
The EU member states representatives have been closely involved to ensure wide-public participation process, via the AI Board. The AI Office also invited other public bodies and agencies from all over the world working on risk assessment and mitigation for general-purpose AI models to the Plenary as observers.
The AI Office played a pivotal role throughout the process: e.g. by facilitating the drawing-up, coordinating the discussions and documenting the outcomes, the AI Office guaranteed that the Chairs, Vice-Chairs, and all the participants in the Plenary had access to the information and can contribute meaningfully. This oversight helps maintain an open and collaborative environment, fostering trust and accountability in the development of the Code of Practice.
Performance
GPAI models can perform a wide range of tasks; besides, they are becoming the basis for many AI systems in the European states: some of these models could carry systemic risks if widely used in Europe.
To ensure safe and transparent AI for the member states, the AI legal act introduces rules for providers of such models, including transparency and copyright-related rules. It is suggested that for models that may carry systemic risks, providers should assess and mitigate these risks.
The AI Act rules on general-purpose AI shall apply from the beginning of August 2025; the AI Office has been facilitating the drawing-up of a Code of Practice to detail out these rules.
The following three staying separately code’s chapters: transparency, safety of copyright and security.
Thus, chapters on transparency and copyright offer all providers of general-purpose AI models demonstrating compliance with their obligations under “obligations for providers of Gen-purpose AI models” (art. 53 AI regulation on “harmonized rules” nr.1684 from June 2024).
Whereas the chapters on safety and security are only relevant to the small number of providers of the most advanced models, those that are subject to the AI Act’s obligations for providers of general-purpose AI models with systemic risk (under art.55 of the AI Act).
Template for the summary of training data
In parallel to the code of practice process, the AI Office is also developing a template on the sufficiently detailed summary of training data that general-purpose AI model providers are required to make public. i.e. “draw up and make publicly available a sufficient detailed summary about the content used for training of the general-purpose AI model” (art. 53, 1d).
The AI Office would collect input from stakeholders during a broader multi-stakeholder consultation on general-purpose AI, where more than 430 responses were provided from a wide range of stakeholders.
The template for the summary of training data is closely linked to the providers’ obligations in relation to transparency and copyright that have been detailed out in the code of practice. Given the close interlinkages of the two processes and the participation of all interested stakeholder groups with more than 1000 participants in the code of practice drafting, the AI Office has presented its preliminary ideas and allowed the participants to the Code to provide additional feedback on the preliminary structure and elements of the template.
This issue has been also discussed with the EU member states’ representatives in the AI Board subgroup and the European Parliament before the Commission adopted the template in the second quarter of 2025.