The Society of Authors (“SoA”) has written to Chief Executives of AI developers, putting them “on express notice” that the SOA’s over 12,500 members do not authorise the use of their works for the development of AI systems unless they have specifically agreed licensing arrangements. Although it does not expressly say so, this letter is presumably intended to function, to the extent required in the EU, as the so-called text and data mining ‘opt-out’ under Article 4 of the DSM Copyright Directive, as well as delivering the necessary warning in the UK and elsewhere.

The SoA is the largest trade union of its type, representing – among others – writers, scriptwriters, illustrators, and literary translators, and is the latest group to express concern about the unauthorised use of copyright-protected works to develop generative AI systems. On 2 May 2024, an Extraordinary General Meeting was convened at which members of the SoA voted in favour of a resolution to write to tech companies on this subject.

The letter cites various provisions in the Copyright, Designs and Patents Act 1988 (“CDPA”), arguing that “the use of copyright works as input to train and develop an AI model without the licence or consent of the rightsholder amounts to copyright infringement (s. 16(2) and s.17 CDPA) and the copyright is further infringed where the AI model is used to generate a work (“the Infringing Work”) which reproduces the whole or a substantial part of the copyright work (s.16(2); s.17; s.18; and s.20 CDPA).”

Furthermore, the SoA argues that the training, operation and development of AI models using copyright-protected works without permission does not fall within any of the ‘permitted acts’ under the CDPA, and that no defences would be available to AI developers. As such, the letter states that “our members have instructed us to put you on express notice that they do not authorise or otherwise grant permission for the use of any of their copyright-protected works in relation to, without limitation, the training; development; or operation of AI models (including the generation of Infringing Works), by large language models or other generative AI models, unless they have first specifically agreed licensing arrangements for the use of their works.”

The letter points out at the outset that the SoA’s members urge AI developers “to agree terms on a commercial basis with respective rightsholders given that licensing opportunities exist and are being developed”. Until that time, it calls on developers to do the following:

1. identify the works which have been used to date to develop its AI model;
2. undertake to make suitably detailed requests for permission to use any of the SoA’s members’ works in the future;
3. undertake that, before using any copyright-protected works, they will first obtain permission from the relevant rightsholder;
4. undertake to pay appropriate remuneration for all uses of copyright-protected works, both past and future;
5. undertake to give appropriate attribution of the author of the work in all cases; and
6. undertake that, on request (whether general or in relation to a specific work), they will remove any work which has been used without permission from their systems and will provide evidence of compliance.

In the press release accompanying the letter, the SoA states that “tech companies must agree licences and seek agreement from authors for all past and future uses… There are options available to ensure that generative AI and creators can coexist. This is vital not only for creators, but also for successful generative AI systems, which depend on human authored work to function and to evolve.”

To read the letter in full, click here.