Open Future proposes EU fund for AI training remuneration
Open Future has proposed a collective remuneration mechanism that would require companies supplying access to certain AI models in the European Union to contribute part of their EU revenue to a shared fund. The proposal is presented in a discussion paper and as the organisation’s contribution to a European Commission consultation on copyright and European creativity.
- Open Future proposes a single EU-level fund for AI training remuneration.
- Payments would be based on EU revenue from commercially supplied model access.
- Research and open-source model development would fall outside the obligation.
- Creators, rights holders and public content stewards would share the fund.
- The paper leaves several design questions open for discussion.
A collective alternative to individual licensing
Open Future’s paper, Sharing the Value of AI training, addresses the tension between two goals identified by the European Commission: improving remuneration for creators while making it easier for generative AI providers to access copyright-protected content.
According to Open Future, individual licensing cannot reach most of the material used to train AI models. The organisation says licensing tends to take place between the largest AI developers and the largest rights holders, leaving out individual creators, small rights holders and smaller AI developers. Its proposal therefore shifts the focus from transaction-by-transaction licensing to collective remuneration.
The paper builds on Open Future’s Beyond AI and Copyright, published in June 2025, and follows Mistral’s call for a European content levy earlier in 2026.
How the proposed mechanism would work
The obligation would be triggered when a company commercially supplies access on the EU market to AI models trained on publicly available content. It would apply regardless of where the company is established or where the model was trained. Companies would pay a percentage of their EU revenue from supplying access to a single EU-level fund.
In return, training on publicly available content would no longer be subject to individual reservations under Article 4 of the CDSM Directive. Remuneration would be provided through a statutory claim that cannot be waived and would be managed collectively.
Research, along with the development and release of models under open-source licences, would fall outside the obligation. The obligation would arise only when such a model is commercially supplied. Content behind paywalls or other access controls would remain outside the mechanism and available for licensing. Uses at the inference stage, including retrieval-augmented generation and grounding, would not be affected.
Fund distribution and unresolved questions
The larger share of the fund would go to creators and rights holders through collective management organisations, with a guaranteed portion reserved for authors and performers. A significant fixed share would go to stewards of publicly available content, including libraries, archives, museums, open knowledge projects and public service media.
Open Future stresses that the paper is not legislative text. Instead, it sets out requirements for a possible intervention and examines six elements, including the trigger, calculation of payments, distribution of proceeds, interaction with existing copyright rules and legal form.
Questions remain over the boundary for open models, the unit of use for deemed model access and the size of the fund’s shares. Open Future describes the proposal as an invitation to discuss a European approach to AI and copyright. The organisation says it developed the proposal with stakeholders including Mistral and Black Forest Labs, while retaining sole responsibility for its content.
Open Future: Sharing the Value of AI training
