The UK is “on the cusp of an AI licensing boom that could cement its status as a global creative and tech superpower”, according to record industry trade group BPI, but only if the government “acts now to maintain intellectual property and mandate greater transparency by AI developers”.
In a new report, the BPI notes that the UK is both the third biggest recorded music market in the world (after the US and Japan) and the third biggest AI market in the world (after the US and China).
And while the debate over the copyright obligations of AI companies has been “polarised to date”, the BPI’s report optimistically concludes that “there is potential to change that and to find a way forward that serves both industries”.
Though that will require AI companies seeking licences for any existing content they use, including music, when training generative AI models. The BPI report notes that some music AI companies are now doing that and have entered into licensing deals with the majors and the indie label repping Merlin.
Meanwhile, 16% of indie labels surveyed by the BPI say they are already exploring licensing partnerships with AI companies, while 77% said they would be open to licensing their catalogues for “ethical use by AI”.
The good old general public also believes AI companies should secure licences, the BPI says. It surveyed 3000 consumers of which 74% said that rightsholders and artists should be paid if their work is used to train AI music systems.
But there, of course, a whole load of AI companies that are making use of existing content without securing any licences. Which means, the BPI says, “unlocking the AI licensing opportunity at scale” will require intervention from the government. But, “the policy conversation must shift away from unnecessary changes to copyright law and focus instead on transparency, labelling and the sovereignty of UK law”.
Beyond the stats and updates on the music industry’s deals with AI businesses, the BPI’s report sets out the trade body’s lobbying priorities regarding music and AI, now that the UK government has abandoned its controversial plan to introduce a new copyright exception that would have benefited the AI sector.
Since dropping that plan, ministers have embarked on yet another round of consultation with creators, rightsholders and AI companies.
And while we know what that consultation will consider - labelling of AI-generated content; digital replicas; creator control and transparency; and the impact of AI on independent creatives - we don’t really know what new regulation or legislation the government may ultimately propose.
The BPI sets out its priorities in the new report. That includes ‘mandating transparency requirements’, so that AI companies are transparent about what content they used to train their models, and AI-generated content is clearly labelled.
The trade body notes that “a number of industry-led initiatives are already in place to enable labelling”, but adds “it would be helpful for the government to create a requirement in law or regulation for the labelling of AI-generated content”.
But perhaps more importantly, the BPI wants the government to “assert the sovereignty of UK law”, which basically means that any AI company making its models available in the UK should have trained those models according to UK copyright law.
Under the UK copyright system, AI companies almost certainly do need licences for any content they use when training their models and a failure to secure those licences is copyright infringement. But in some other countries there are copyright exceptions - or in the US the ‘fair use’ principle - that might allow AI training without getting permission from rightsholders.
So what happens if an AI model is actually trained in one of those countries but then commercialised in the UK? That question was posed in the UK courts last year in a legal battle between Getty Images and Stability AI, which had used Getty’s images when training its image-generating AI model, but that training happened in the US.
Getty tried to argue there was still a case for copyright infringement under UK law, relying on copyright rules that relate to the importing of infringing goods, but those arguments did not succeed in court.
So maybe copyright law needs to be changed to clarify things in this domain. The BPI says, “it should be made clear that all AI models deployed in the UK must comply with UK law irrespective of where they have been trained”, adding “this would stimulate commercial licensing, prevent circumvention of UK law and ensure a more level playing field for the UK’s ethical AI companies”.
While the BPI’s demands of government - for transparency and the ‘sovereignty of UK law’ - will be supported by the wider music community, artists and songwriters have other concerns too, not mentioned by the record industry trade group, where they also would like the government to intervene.
As the major labels start to do licensing deals with AI companies, that creates its own concerns for artists and songwriters, who are often left in the dark about how those deals will work, when creator consent will be sought by labels, and how the money will be shared.
Which means organisations representing music creators will be asking for even more intervention from the government, to ensure that any AI licensing boom benefits artists and songwriters as well as their label and publisher partners.