The music industry has welcomed - albeit cautiously - yesterday’s UK government announcement on AI and copyright, in which ministers scrapped plans for a controversial ‘text and data mining’ - or TDM - copyright exception that would benefit AI companies.
However, the downside is that the government now basically has no plan at all on AI and copyright, which means there remains concern that a proposal for another AI copyright exception could rear its head in the future.
The announcement came on the same day Sony Music’s Dennis Kooker revealed that the major has now asked streaming platforms to remove more than 135,000 tracks featuring unapproved AI voice clones of Sony-signed artists. Meanwhile, in the US, yet another lawsuit was filed by a music company accusing an AI business of copyright infringement, with BMG suing Anthropic.
The UK government first proposed the new TDM exception in late 2024, presenting it as a preferred approach when opening up a big old consultation on copyright and AI.
The exception would have allowed tech companies to use existing content when training AI models, including music, without getting anyone’s permission, except where a rights owner had ‘reserved their rights’, so formally opted-out of the exception through some technical process.
That proposal was immediately slammed by the music and wider creative industries, prompting public protests and relentless lobbying. As a result, ministers spent much of the last year backtracking before finally confirming yesterday that a TDM exception with opt-out was no longer their preferred option
Welcoming that development, Roberto Neri, CEO of The Ivors Academy, says: “A new text and data mining exception would have fundamentally weakened copyright protections for creators”.
“Avoiding this has been achieved by making the case to policymakers and sustained engagement across the creative industries”, he adds. “The UK must now deliver a framework where AI companies license creative works with authorisation, fair remuneration and transparency”.
Meanwhile Sophie Jones, CSO at record industry trade group BPI, says, “We welcome the government taking its previously preferred option of a TDM with opt out off the table - any new exceptions to copyright for AI training would be deeply damaging to the music industry and wider creative economy”.
However, music industry groups have also noted that, while yesterday’s announcement was good news, the UK government has not been as bold as their Australian counterpart, which last October ruled out introducing any kind of TDM exception. UK ministers have killed the specific kind of TDM exception they previously said they preferred, but could still come back with an alternative exception in the future.
AI expert and campaigner Ed Newton Rex put it like this in a social media post. “The good news for creatives is that the government formally dropped its preferred option of a broad copyright exception with an opt-out. The bad news is that weakening copyright law is very much still on the table”. Ministers are
“explicit”, he adds, “that they are still considering other forms of copyright exception”.
UK Music CEO Tom Kiehl, while welcoming yesterday’s announcement, also honed in on this point, urging the government to “go further and rule out resurrecting this plan throughout their period in office”. It remains “vitally important” he added, “that the government does not now consider any kind of alternative copyright exception that would negatively impact creators and rightsholders in the music industry”.
With the government swapping its unpopular plan around copyright and AI for basically no plan at all, ministers will now instigate yet another round of research and consultation on the off chance they stumble across a new less controversial plan in the process.
That includes a “taskforce” to consider labelling AI-generated content, a “review” of the mechanisms available for creators to control their works online, and a “consultation” on digital replicas.
Digital replicas are another key area of concern for the music industry, which wants to ensure that artists have control over their voice and likeness in the context of AI. This takes us beyond copyright and into legal principles such as publicity and personality rights, which don’t currently exist in the UK.
Sony Music’s President of Global Digital Business Dennis Kooker discussed the negative impact of voice clones and deepfakes during the launch of the IFPI ‘Global Music Report’ in London yesterday, which is where he revealed that the major has now requested the removal of more than 135,000 tracks that featured unapproved digital replicas of its artists.
According to the BBC, he explained that these voice clone tracks cause “direct commercial harm to legitimate recording artists”, and often deliberately target artists when they are busy promoting new music.
“That is when deepfakes are at their worst”, he said, “building off and benefiting from the demand the artist has created [and] ultimately detracting from what the artist is trying to accomplish. In the worst cases, they potentially damage a release campaign or tarnish the reputation of an artist”.
There remains much debate over the best ways for artists to protect their voice and likeness, both legally and logistically speaking.
Proposed new laws that would create a new ‘digital replica right’ in the US are currently working their way through Congress and could be passed this summer. Meanwhile in Europe, formal proposals for new digital replica protections are in motion in Denmark and the Netherlands. Presumably the UK government’s digital replicas consultation will consider something similar here.
Meanwhile, back in the US - where the copyright obligations of AI companies are at the centre of dozens of active lawsuits - BMG has now officially joined the AI litigation party. It has sued Anthropic, which is accused of copying lyrics controlled by the music company when training its Claude AI model without securing the necessary permissions.
Anthropic has already been sued by Universal Music Publishing, Concord and ABKCO who make similar allegations. AI companies like Anthropic continue to argue that AI training constitutes ‘fair use’ under US law, which means it doesn’t need permission to make use of existing works.
However, in a legal dispute involving book authors, a judge said AI training definitely wasn’t fair use if an AI company started with pirated copies of copyright protected works. Which Anthropic usually did. Which is presumably why BMG thinks it’s got a decent case here.
“Anthropic’s infringement of BMG’s lyrics causes significant and irreparable harm to BMG and the songwriters it represents, including those who have written the soundtrack to our lives”, the company’s lawsuit states.
It then adds, “Anthropic has profited and continues to profit from its unauthorised use of BMG’s copyrighted musical compositions while the authors of these works are left to watch their creative talents exploited without any compensation or acknowledgment”.