EU

The European Commission is closing its consultation on a draft implementing regulation on detailed arrangements for certain proceedings under the AI Act.

The draft states that it lays down detailed arrangements and conditions for the evaluation of general-purpose AI models under Article 92, including procedures for involving independent experts and selecting them. It also lays down detailed arrangements and procedural safeguards for proceedings in view of the possible adoption of decisions under Article 101 of Regulation (EU) 2024/1689.

Under Article 2, a European Commission decision requesting access to a general-purpose AI model would have to specify the technical means, components, and conditions by which the provider must provide that access. The draft states that access may include APIs, internal access, source code, model weights, access to the infrastructure used to host the model, access to inspect and modify system state, and all levels of access granted to the provider’s own employees.

The draft also states that the European Commission may require a provider to disable and remove logging measures that could track or record the Commission’s access, to the extent necessary to ensure the integrity and confidentiality of the evaluation process. Providers who requested access would have to provide it in a timely and effective manner.

Regarding independent experts, the draft states that the European Commission must take into account factors such as shared ownership, governance, management, personnel, resources, and contractual relationships when assessing independence. It also states that appointed experts must remain independent throughout their appointment and that the confidentiality, integrity, and availability of sensitive information must be protected.

For proceedings that may lead to fines, the draft states that the European Commission may initiate proceedings against relevant conduct by providers of general-purpose AI models. It also states that the Commission may, by decision, order interim measures on grounds of urgency due to a risk of serious damage to health, safety requirements, or other grounds relating to the public interest covered by Regulation (EU) 2024/1689, including preventing a general-purpose AI model from being made available on the market, based on a prima facie finding of an infringement.

Procedural safeguards include written observations on preliminary findings, with a time limit of at least 14 days set by the European Commission, and rules governing access to the file. The draft states that the addressee may obtain access to documents mentioned in the preliminary findings, subject to redactions protecting business secrets and other confidential information, while broader access may be granted under terms of disclosure set by the Commission.

The annex sets format and length requirements for written observations submitted under Article 7. It states that observations must be submitted in a format that allows electronic processing, digitisation, and character recognition, and sets requirements for page format, font, spacing, margins, and numbering. Written observations must not exceed 50 pages, while annexes do not count towards that limit if they have a purely evidential and instrumental function and are proportionate in number and length.

The draft also lays down limitation periods for the imposition and enforcement of penalties, rules on the beginning and setting of time periods, and provisions on the transmission and receipt of information. It states that documents transmitted by digital means must use at least one qualified electronic signature and that, for real-time or near real-time information shared through APIs or equivalent means, the European Commission will define the methods and duration of that sharing.

The regulation states that it would enter into force on the twentieth day following its publication in the Official Journal of the European Union.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!