27 June 2023

Machine learning allows computers to generate new works and processes without human interference, such as software, photographs, paintings and books. If this happens, who owns the copyright?

Under the Copyright Designs and Patents Act 1988 (CDPA) the first owner of copyright will be the author or in certain cases their employer. The CDPA states that for a computer-generated literary, dramatic, musical or artistic work, the author is the person by whom the arrangements necessary for the creation of the work are undertaken. “Computer-generated” means generated by computer with no human author.

These definitions are ambiguous. AI is not (yet) a legal person or entity: it cannot enter into a contract or act as an agent for another. The words “undertaking necessary arrangements” for the creation of the work where “there is no human author” still need to be tested by case law. So if the developers of an AI system believe that it could generate new works, any contract for its development should make it very clear who would own the copyright in any new AI-generated copyright works, and who is therefore able to exploit them commercially by, for example, assigning/selling or licensing them to others.

There are other grey areas. A human being can store a photograph in their memory without infringing copyright. But what happens when a copyright work is copied, stored or used in an AI “brain” to enable machine-learning to occur?

For example, a developer might upload the works of author John Grisham into an AI system in order to ask the AI system to write a Grisham-style novel. Without a licence from the copyright holder, the act of uploading those copyright works, to “train” the AI system, might itself be a breach of copyright.

A substantial part of a work needs to be copied to prove infringement. However, section 29 of the CDPA allows you to make of copies of text and data analysis for non-commercial research. There is clearly scope for abuse: for example, you could get an AI system to produce 50 different AI novels, all based on original copyrighted works, for “non-commercial research”. If you then used the AI-produced works to create more works for commercial use, it is possible you might escape claims of copyright infringement provided that no AI work was a substantial copy of an existing copyright work.

Technology and the law rarely develop at exactly the same pace, so the law will need to keep playing catch-up. If you need any copyright or intellectual property advice, or have questions about AI, please contact Brian Levine on (01935) 846 258 or email him at