The European Union has launched a consultation aimed at establishing rules for providers of general-purpose AI models (GPAIs) such as Anthropic, Google, Microsoft, and OpenAI, as part of the AI Act—a risk-based framework for regulating artificial intelligence applications. Lawmakers emphasize that the forthcoming Code of Practice will help promote "trustworthy" GPAIs by offering developers clear guidelines to meet their legal obligations.
The AI Act was adopted earlier this year and is set to take effect on August 1. Compliance will be rolled out in phases, with the Codes of Practice scheduled to be enacted by April 2025, providing the EU with time to develop comprehensive guidance.
The European Commission is encouraging input from GPAI providers operating within the EU, as well as businesses, civil society representatives, rights holders, and academics. “This consultation offers all stakeholders the chance to voice their opinions regarding the initial Code of Practice, which will clarify the rules for general-purpose AI model providers,” the Commission stated. “The feedback will also shape the AI Office's work, particularly regarding the template for summarizing content used in training general-purpose AI models and the corresponding guidance.”
The consultation consists of a questionnaire divided into three sections: the first focuses on transparency and copyright requirements for GPAIs; the second addresses risk taxonomy, assessment, and mitigation for GPAIs considered to have systemic risk (defined in the AI Act as models trained beyond a specified computational threshold); the third section involves the review and monitoring of the Codes of Practice for GPAIs.
An initial draft of the Code will be created based on the insights gathered from responses to targeted questions. Participants in the consultation will be able to influence the framework the AI Office develops to help GPAI providers meet the legal requirement to summarize their model training content. It will be intriguing to see how comprehensive this template ultimately becomes.
For further details on the consultation, visit the provided link. The deadline for submissions is September 10, 2024, at 6 p.m. CET.
The EU is also seeking expressions of interest from stakeholders interested in contributing to the drafting of the Code through virtual meetings, organized into four working groups. A collaborative drafting process will ensure comprehensive guidance.
The AI Office invites “eligible general-purpose AI model providers, downstream providers, civil society organizations, rights holders, academia, and independent experts to express interest in helping to shape the Code of Practice.”
The deadline for submitting an expression of interest for participation in the drafting process is August 25, 2024, at 6 p.m. CET.
Additionally, GPAI providers will have opportunities to engage in workshops with the plenary meeting chairs and vice chairs. According to the AI Office, these workshops aim to “inform each iterative drafting round, along with plenary participation.”
“The AI Office will ensure transparency during these discussions by creating meeting minutes and sharing them with all plenary participants,” they explained.
The AI Office will appoint meeting chairs and vice chairs and is currently accepting applications from “interested independent experts” for these pivotal roles.
The call for participation and the consultation process respond to concerns that civil society organizations might be marginalized in the drafting phase. Earlier this month, Euractiv reported that the Commission was considering using consultancy firms for the Code, raising worries about potential bias favoring AI industry giants.
In response, the Commission expressed its commitment to inclusivity: “All interested parties are encouraged to participate,” it stated on Tuesday. “The AI Office seeks submissions from a diverse group of stakeholders, including academia, independent experts, industry players like general-purpose AI model providers, civil society organizations, rights holders, and public authorities.”
Ranking
