Introduction
The European Commission’s Artificial Intelligence Office has initiated a comprehensive consultation process to gather input on future guidelines for defining AI systems and identifying prohibited practices under the EU AI Act. This initiative aims to enhance the relevance of the guidelines and clarify the practical implications of the Act, which is recognized as the first comprehensive AI law globally [3]. The consultation involves a diverse array of stakeholders [10], including businesses [1] [2], research institutions [2] [5] [7] [11], non-governmental organizations [2], and individual citizens [2].
Description
The European Commission’s Artificial Intelligence Office has launched a multi-stakeholder consultation process aimed at gathering input on future guidelines for defining AI systems and identifying prohibited practices under the EU AI Act. This initiative [2] [5] [6] [7] [8] [9] [10] [11], which began on 13 November 2024 and will conclude on 11 December 2024, invites contributions from a diverse array of participants, including businesses [1] [2], research institutions [2] [5] [7] [11], non-governmental organizations [2], and individual citizens [2]. The goal is to enhance the relevance of the guidelines and clarify the practical implications and use cases of the AI Act, which is recognized as the first comprehensive AI law globally [3].
The EU AI Act introduces a risk-based regulatory framework for AI systems [3], set to be implemented in phases starting from 2 February 2025 [3]. This framework is designed to assist national authorities [5] [11], as well as providers and deployers of AI [5] [11], in adhering to the Act’s regulations [5] [7] [11]. Key players in the AI industry, including OpenAI [2] [5] [7] [10] [11], Google [4], Meta [4], and Anthropic [4], will need to comply with the forthcoming Code of Practice to avoid enforcement actions. Responses collected during the consultation will be analyzed to inform the development of the guidelines, which are expected to be finalized and published in early 2025 [10]. The Commission aims to create clear and enforceable provisions that support safe and ethical AI practices across the EU [10], ultimately aiding businesses and public authorities in navigating the new regulatory landscape [10].
Legal experts anticipate that the consultation will help clarify definitions of AI systems [10], which include machine-based systems designed to operate autonomously and adapt post-deployment [3], generating outputs such as predictions or decisions based on input data [3]. This clarification will assist companies in determining whether their technologies qualify as AI and ensure consistent enforcement and compliance with the Act. The Code of Practice will guide general-purpose AI (GPAI) providers in meeting their obligations under the EU AI Act, allowing for deviations if compliance can be demonstrated through alternative measures [4].
In parallel, independent experts have released the first draft of the General-Purpose AI Code of Practice [8], which will be discussed with approximately 1,000 stakeholders through an iterative drafting process that began on 30 September [8]. This process will continue through four rounds until April 2025 [8], allowing for comprehensive feedback to shape the evolution of the draft. The initial draft [8], developed by appointed experts from thematic working groups [8], incorporates input from providers of general-purpose AI models and considers international approaches [8].
The draft outlines guiding principles and objectives [8], aiming to clarify the potential structure and content of the final Code [8]. It establishes clear objectives [8], measures [1] [4] [8], and key performance indicators (KPIs) to guide the development and deployment of trustworthy general-purpose AI models [8], addressing transparency and copyright compliance for AI model providers [8]. Notably, companies will be required to disclose information about the web crawlers used in training their models [1], which is crucial for copyright holders [1]. The draft also includes a taxonomy of systemic risks and mitigation measures for advanced models that may pose systemic risks [8], including prohibitions on manipulative practices, unauthorized facial image scraping [3], exploitation of vulnerable individuals [3], and detrimental categorization of people [3]. The most powerful GPAIs [4], defined as models trained with over 10^25 FLOPs of computing power [1] [4], will be subject to EU AI Act rules aimed at mitigating systemic risk [4].
Next week [8], the draft of the General-Purpose AI Code will be discussed in dedicated working group meetings with stakeholders, representatives from EU Member States, and observers [8], allowing participants to provide verbal feedback on recent drafting progress [8]. A draft has been shared on a dedicated platform [8], with a two-week window for written feedback [8]. Ongoing legal challenges against AI firms regarding the unlawful processing of copyrighted data highlight the importance of compliance requirements related to information provisions [4], particularly concerning copyrighted material [4].
The finalized guidelines for the EU AI Act are anticipated to provide practical guidance to businesses, regulators [2] [3] [4] [5] [6] [7] [10] [11], and other stakeholders in fulfilling their responsibilities under the Act [6]. The drafting principles emphasize proportionality of measures to risks [8], consideration of provider size [8], and simplified compliance options for SMEs and start-ups [8], while also reflecting exemptions for open-source model providers and balancing clear requirements with flexibility to adapt to technological advancements [8]. Transparency requirements for GPAI makers will take effect on 1 August 2025 [4], while those classified as having “systemic risk” must adhere to risk assessment and mitigation requirements by 1 August 2027 [4]. The AI Office is mandated to develop Codes of Practice by May 2025 [3], while the Commission will provide further clarity through guidelines and Delegated Acts [3], ensuring that stakeholders are well-informed about the implications of the legislation on their AI system uses. Non-compliance with the AI Act can result in significant penalties [1], including fines up to €35 million or seven percent of global annual profits [1], whichever is greater [1]. This comprehensive approach underscores the EU’s commitment to fostering innovation while safeguarding public interests in the realm of AI technologies.
Conclusion
The consultation process initiated by the European Commission’s Artificial Intelligence Office is a significant step towards establishing a robust regulatory framework for AI systems under the EU AI Act. By involving a wide range of stakeholders [2], the initiative aims to ensure that the guidelines are relevant and practical, supporting safe and ethical AI practices across the EU [10]. The anticipated guidelines and Codes of Practice will provide essential guidance for businesses, regulators [2] [3] [4] [5] [6] [7] [10] [11], and other stakeholders [2] [5] [6] [10] [11], helping them navigate the new regulatory landscape [10]. The EU’s commitment to balancing innovation with public interest protection is evident in its comprehensive approach, which includes significant penalties for non-compliance, thereby reinforcing the importance of adherence to the Act.
References
[1] https://www.engadget.com/ai/the-eu-publishes-the-first-draft-of-regulatory-guidance-for-general-purpose-ai-models-223447394.html
[2] https://iistl.blog/2024/11/22/a-step-towards-responsible-ai-in-europe-consultation-and-draft-code-of-practice-unveiled/
[3] https://www.pinsentmasons.com/en-gb/out-law/news/eu-ai-act-guidelines-scope-prohibitions-come-early-2025
[4] https://techcrunch.com/2024/11/14/eu-ai-act-draft-guidance-for-general-purpose-ais-shows-first-steps-for-big-ai-to-comply/
[5] https://digital-strategy.ec.europa.eu/en/news/commission-launches-consultation-ai-act-prohibitions-and-ai-system-definition
[6] https://ai-regulation.com/other_tags/ai-office/
[7] https://healthmanagement.org/c/it/PressRelease/commission-launches-consultation-on-ai-act-prohibitions-and-ai-system-definition
[8] https://digital-strategy.ec.europa.eu/en/library/first-draft-general-purpose-ai-code-practice-published-written-independent-experts
[9] https://digitalpolicyalert.org/change/11817-commission-guidelines-on-the-application-of-the-definition-of-an-ai-system-and-the-prohibited-ai-practices-established-in-the-ai-act
[10] https://ai-regulation.com/commission-consults-ai-act-definitions-prohibitions/
[11] https://eoc.org.cy/commission-launches-consultation-on-ai-act-prohibitions-and-ai-system-definition/