If you work in the creative industries, you might be concerned about the increasing prevalence of AI tools and their capabilities. An AI model can generate useful text, images, audio and even videos with relative ease – if it has been trained correctly, you give it the right prompt, and it doesn’t hallucinate.
The UK Government is currently asking what people think about AI and copyright, in a drive to figure out how best to support (read: leverage) the AI sector and the creative industries. If you are interested in responding to the consultation, then you have until 25 February 2025.
Should the AI sector be given freedom to use other people’s creations in the pursuit of ever more sophisticated AI models? Should the rights of human authors be paramount? What is the right balance?
Sir Paul McCartney (who you might know from somewhere) weighed in with some choice words about the UK government’s intended approach of an “opt out” system. That is, a system where you would have to take an active step to stop the AI sector using your stories, photos, artworks, music, videos, and so on. Failing to opt out would mean that it would all be fair game – the AI sector would be able to continue to train AI models to generate new works which are effectively based (to some extent) on what you previously created. Somehow I doubt Sir Paul will be treating us to a rendition of With a Little Help from AI Friends of his own volition any time soon.
Of course, AI models don’t strictly ‘copy’ like a human might. The initial version of an AI model has to be trained on a huge set of input data, told what is right and what is wrong over many iterations, and gradually gets nudged into a position where it can provide a useful desired output.
One example that went viral in 2023 was the AI-generated song that mimicked Drake and The Weeknd. But the song was not a copy – it was not a reproduction of something that was protected by copyright. It was a brand-new song that convincingly emulated their voices. That being said, the AI model only reached the stage where it could effectively imitate genuine artists after being specifically trained on the artists’ works, and the money generated from the work isn’t going to the AI model. Shouldn’t the artists be entitled to some benefit?
The law tends to lag behind technological advances, and AI is no exception. Copyright does not protect a person’s voice, at least not currently. The law of passing off might help in some cases, but only if there is misrepresentation, i.e. the public is being deceived as to the origin of the work, which wasn’t the case for that song. So, is a new or modified law now required to enshrine rights for things that could not, until recently, be copied? Or are our existing laws enough?
At the moment, AI-related tools are already being developed on an accelerated trajectory, and it makes sense for those tools to be built and used for the benefit of everyone. But I think that the existing rights of authors, artists and any other creator must be respected in the process of doing so. After all, AI is not itself being granted any rights – rather, it is the people and companies in the AI sector that could end up being given preferential treatment at the expense of the vast number of people creating new stories and art in all its various forms. How many authors and artists would really be able to effectively opt out every single new work that they create and make available to the world, and what would happen if they didn’t opt out quickly enough? To quote Sir Paul: “We’re the people, you’re the government! You’re supposed to protect us. That’s your job.”
Given that it is becoming increasingly difficult to identify whether something is AI-generated or not (finger trouble aside), it is arguably more important than ever to properly legislate and regulate how the AI sector operates in the first place… which leads me nicely back to the UK government’s consultation. I hope Sir Paul doesn’t mind that I shall summarise it as: Come Together and Help!
Author
