Expert Policy Proposal
The AI Act and General Purpose AI
Key Recommendations to inform EU's AI Act Negotiations regarding General Purpose AI
As the European Union's Artificial Intelligence Act enters its final phase of negotiations, AlgorithmWatch and the AI, Media, and Democracy Lab of the University of Amsterdam publish key policy proposals for governing General Purpose AI (GPAI) and Generative AI (GenAI). The recommendations were developed with substantive input by leading experts in the field.
The paper highlights actionable suggestions that should be considered during the ongoing trilogue negotiations.
The policy briefing delves into five critical areas for governing GPAI and GenAI:
- Clarifying definitions
- Addressing complexity, scale and power asymmetries
- Avoiding accountability gaps along the value chain
- Ensuring democratic oversight
- Complementing other regulations
Accountability gaps in GPAI governance are a real concern. This paper highlights the importance of a meaningful division of responsibilities, robust rights for affected individuals and coherent and technology-neutral definitions. We encourage all parties to consider these recommendations during the trilogue negotiations on the AI Act. By addressing these key issues, the EU can live up to its commitment to responsible AI governance and championing fundamental rights.
Correction notice: On September 15, we updated the paper by deleting the recommendation on copyright law in section 5, "Complementing other Regulations", and updating the list of authors. Please only refer to the file PolicyBrief_GPAI_AW-updated1509-02.pdf as the authoritative version.
The workshop and the preparation of the policy paper was supported by the Alfred Landecker Foundation as part of the project Auditing Algorithms for Systemic Risks.
Read more on our policy & advocacy work on the Artificial Intelligence Act