Regulation of artificial intelligence and copyright: What can legislation influence?

Q: What is an extended collective licence?

An extended collective licence refers to a rights management system as defined in the Copyright Act. An extended collective licence allows for a licence issued by an organisation that represents copyright holders to be extended to cover the works of authors and copyright holders not represented by the organisation. The system enables copyright organisations to grant comprehensive licences for the use of works, enabling companies and various organisations to obtain legally certain and comprehensive licences through a centralised process, thereby also reducing the associated costs.

Q: What does it mean that the EU should promote the licensing of copyrighted content for use by AI?

The EU should promote practical solutions that facilitate the acquisition of collective licences and the responsible use of copyrighted content in the training of artificial intelligence. An effective way of achieving this would be to ensure the use of extended collective licences in cases of mass use of works, involving a large number of works or users, across the borders of EU Member States. The use of works for training artificial intelligence should be based, first and foremost, on the authorisation of the copyright holders. Collective licensing arrangements are important so that copyright holders can receive remuneration for the use of their content and AI operators can obtain legally certain licences to use works. This provides sustainable rules for the development of artificial intelligence.

Q: How does the DSM Directive regulate artificial intelligence and copyright?

The DSM Directive, which regulates the EU’s digital single market, does not directly regulate artificial intelligence, as it was drawn up before generative AI applications became widespread. However, the provisions of the Directive relating to text and data mining have come to play a central role in the debate on artificial intelligence. In commercial text and data mining, works that are lawfully available may be used without the copyright holder’s permission, unless the holder has prohibited such use. The training of AI models involves the processing of large amounts of data, but it is legally unclear whether the exception for text and data mining can be applied to the entire operational process of artificial intelligence. The DSM Directive does not take a position on this matter, and there are currently several court cases underway on the issue.

Copyright holders are willing and prepared to grant licences for the use of their works in the development of artificial intelligence, but so far many of the major technology companies have been avoiding negotiations. For this reason, licensing requires legislative backing.

How are artificial intelligence and copyright regulated at the EU level?

A regulatory framework relating to artificial intelligence and copyright does exist, but its practical implementation and interpretation are still evolving. Artificial intelligence is regulated in the EU by the AI Act, which came into force in 2024, with the transparency obligation being a key provision of the Act from the perspective of copyright holders.

The transparency obligation means that providers of general-purpose AI models must adopt operating principles to ensure compliance with EU copyright legislation. For example, AI operators must take into account the prohibitions imposed by copyright holders on the use of their works in commercial text and data mining, which takes place during the training of AI.

Furthermore, AI operators must also publish a sufficiently detailed summary of the data used to train the model, so that it can be assessed whether they are obliged to obtain permission to use the work from the copyright holder.

The AI Act is supplemented by the General-Purpose AI Code of Practice and the Commission’s guidelines. These set out in concrete terms how AI operators can fulfil their obligations regarding copyright and transparency: for example, by publishing their copyright policies, identifying any reservations of rights and providing information on their data mining tools and data sources.

The AI Act also includes an obligation requiring that certain content generated or manipulated by AI must be transparently labelled as having been created or modified by AI. The aim of this obligation is to promote the traceability of the origin of content and to prevent misleading information.

How should regulation at the EU level be developed?

A regulatory framework relating to artificial intelligence and copyright has been established, but the enforcement of its provisions needs to be strengthened. In our view, the widespread adoption of generative artificial intelligence has highlighted the practical challenges of EU copyright regulation in the digital environment. It remains difficult for copyright holders to obtain information on the use of their works, enforce their rights and negotiate licences and remuneration.

The regulation should support a well-functioning licensing market in which the use of works can be agreed upon and appropriate remuneration paid. The EU should promote collective licensing solutions for the mass use of copyrighted content, i.e. situations where there is a large number of works being used or a large number of users of works. This would provide legal certainty for both copyright holders and companies developing artificial intelligence, and would enable the responsible use of high-quality content in the development of artificial intelligence.

The European Parliament has also emphasised that the development of artificial intelligence must be based on respect for copyright, transparency and the possibility of reaching agreements with copyright holders on the use of content.

Which matters can be decided on at the national level?

At the national level, a provision on extended collective licences should be added to the law to cover the use of works in the development of artificial intelligence. With extended collective licences, a copyright organisation approved by the Ministry of Education and Culture may grant licences for the use of works even on behalf of authors other than those it represents, which facilitates the centralised acquisition of licences.

While many AI models have been developed by global companies, these models are also further trained at a local level, which means that national legislation plays a significant role. An extended collective licence would facilitate the development of artificial intelligence with high-quality Finnish works, while respecting copyright.

For example, our benchmark countries, Norway and the Netherlands, have decided to develop their own national language models, which are trained with high-quality data in a responsible manner. This means that the language model is trained by using a wide variety of copyrighted content, for which permission has been obtained from copyright holders and for which remuneration is paid.