Second GPAI Signatory Taskforce meeting - Copyright Chapter

The second meeting of the Signatory Taskforce under the General-Purpose AI (GPAI) Code of Practice - held on 13 March, 2026 - brought together representatives from nearly all Signatories to the Code.

The roundtable discussion focused on two provisions of the Copyright Chapter:

• Measure 1.4 (Mitigating copyright-infringing outputs)
• Measure 1.5 (Designation and functioning of contact points for complaints).

Moderated by the European AI Office, the meeting provided a forum for Signatories to exchange views, share early implementation experiences and best practices, and foster mutual learning.

The discussion on Measure 1.4 highlighted different approaches for mitigating the risk of copyright-infringing outputs. Participants emphasised that a range of technical and organisational practices exist and noted that, to be effective, such measures are typically implemented throughout the model lifecycle - from data curation, model training, evaluations prior to placing on the market, and deployment - with mitigations at both the input and output levels. One participant also presented an approach involving the training of GPAI models on licensed content and the implementation of an attribution mechanism that measures similarities between AI-generated outputs and the training data. This approach aims to enhance transparency and enable fair remuneration for data contributors. The exchange also underlined the importance of ongoing work to identify and implement solutions that achieve effectiveness, transparency, and operational feasibility. It was also noted that model modalities play a role in assessing the effectiveness of the mitigations. Reference was also made to best practices in mitigating ethical risks and developing relevant technical solutions through partnerships with academic research institutions and other civil society organisations. The importance of further developing open-source solutions and their ability to prevent copyright infringements in AI-generated content was also emphasised, as this could benefit the ecosystem as a whole, including smaller model providers.

The exchange on Measure 1.5 focused on initial experiences in establishing and operating contact points and procedures for handling rightsholders complaints. The discussion included organisational arrangements for receiving complaints and ensuring accessible, timely responses. Participants noted that implementation practices may evolve as experience grows and processes are further refined.

The AI Office thanked the participants for sharing their first input on the implementation of these measures and invited them to propose topics of interest to be considered in the preparation of the next meeting of the Signatory Taskforce.

Find further information about:

• General-Purpose AI (GPAI) Code of Practice
• Signatory Taskforce

Related topics