The European AI Office has launched a consultation on trustworthy general-purpose AI models under the AI Act.

As the background document to the consultation explains, the EU AI Act “puts in place effective rules and oversight for general purpose AI model providers”. This includes the development of a Code of Practice which will “reflect the state of the art and duly take into account a diverse set of perspectives”. The consultation launched by the AI Office provides an opportunity for stakeholders to have their say on the topics that will be covered by this Code of Practice.

The consultation covers two broad areas. First concerns “transparency and copyright related rules”. It considers what sort of information and documentation developers of general-purpose AI models should make available to providers of AI systems who intend to integrate general-purpose AI models into their systems, as well as what information should be made available to the AI Office and national competent authorities. On the subject of copyright, the consultation asks what elements should be included in the policy that general-purpose AI models have to put in place “to comply with Union law on copyright and related rights, as required by the AI Act”. Furthermore, it asks what “categories of information sources” should be included in the obligatory summary that developers must provide of the content which is used for training of the general-purpose AI model.

The second area relates to “risk taxonomy, assessment and mitigation”. The consultation sets out that “’systemic risks’ refer to risks that are specific to the high-impact capabilities of general-purpose AI models (matching or exceeding the capabilities of the most advanced general-purpose AI models); have a significant impact on the Union market due to their reach; or are due to actual or reasonably foreseeable negative effects on public health, safety, public security, fundamental rights, or society as a whole, that can be propagated at scale across the value chain”. The Code of Practice intends to establish a ‘risk taxonomy’ of the type and nature of systemic risks, and the consultation seeks views on what form this should take.

Finally, the consultation asks for respondents to set out key performance indicators which will be effective to measure the compliance of participants with the Code of Practice, and how such KPIs can “reflect differences in size and capacity between various providers such as SMEs and start-ups”.

The consultation remains open until 10 September 2024 and can be read here. Separately, stakeholders who wish to participate in the process of drawing up the Code of Practice can express their interest by 25 August here.