Copyright for creatives at risk from AI
A judge’s ruling last week has increased pressure on ministers to act over mounting dilemmas around the impact of technology and artificial intelligence on intellectual property rights.
Singers, songwriters, artists, authors and others want to ensure copyright protection for their creative and original works, while technology companies and innovators want access to that copyrighted material to teach their artificial intelligence systems so they produce the best results.
“Where the balance should be struck between the interests of these opposing factions is of very real societal importance,” Mrs Justice Smith said in the High Court, prompting experts to point to a “growing gap between old copyright law and new technology”.
Smith’s comments came in a ruling on accusations by Getty Images, a US company that creates photographic and video content, against Stability AI, an open-source generative artificial intelligence business, of copyright and trademark infringement.
Getty argued that in “training” its “deep learning” AI model Stable Diffusion — which generates images from text prompts, allowing users to create artwork — the London company scraped and copied about 12 million of the photographic agency’s visual assets without its consent or payment.
During the trial Getty dropped its primary copyright infringement claims because there was no evidence that the training and development of Stable Diffusion took place in the UK.
More broadly, as the judge explains in her ruling, Getty argued that “if creative industries are exploited by innovators such as Stability without regard to the efforts and intellectual property rights of creators, then such exploitation will pose an existential threat to those creative industries for generations to come”.
In the much-anticipated ruling that followed a three-week trial, the judge decided that Stability AI, whose directors include the film-maker James Cameron, did not infringe Getty’s copyright because its image creation model did not store or reproduce any copyright works. She did rule that Stability breached Getty’s trademark in a limited number of cases because some AI-generated images had Getty watermarks.
“The main takeaway is that AI models trained outside the UK, which don’t store or reproduce copyright works, are not exposed to UK copyright when deployed in the UK,” says Alex Shandro, a partner at the law firm A&O Shearman.
Getty, which is pursuing a parallel action in the US, dropped its claim for primary copyright infringement, relating to the act of copying the material, but Shandro says there are still “unanswered questions regarding the legality of training AI systems in the UK and the possible copyright infringement by AI-generated output”.
Nonetheless, Nick Eziefula, a partner at Simkins, says the case “underscores a growing gap between old copyright law and new technology” in the UK.
The ruling prompted lawyers to call on the government to update the law to provide stronger copyright protections against the unauthorised use of work by AI developers. Nick Rose, a partner at Fieldfisher who acted for Getty Images, says the judgment “sets a high bar for proving secondary infringement” — importing or selling material — in which training systems on reproductions of copyrighted data without authorisation is not enough to found a successful claim.
‘There’s a growing gap between old copyright law and new technology’
“While this is an evolving area of law, perhaps it is time to consider the policy implications of this ruling and whether the current legislation is fit for purpose,” he suggests.
The UK’s copyright law was written in the late 1980s when such problems did not exist, says Rebecca Newman, a director at Addleshaw Goddard, as she echoed the call for ministers “urgently to consider how UK copyright law should apply in a machine learning context”.
Although the evidence before the court means that this is a relatively narrow judgment, it is “hugely significant” for the principles it establishes, Gareth Dickson, from Mishcon de Reya, says. Aside from the ruling that Stability AI is not “secondarily liable” for its importation and offering of images created abroad using copyrighted material, he says the trademark ruling was “highly significant”.
Dickson says: “Many trademark owners may take from this that, with better evidence and the correct specifications in their registrations, there are viable claims against providers of AI models for trademark infringement claims to be brought, where outputs contain identical or similar signs.”
In addition, the judgment confirms that an intangible electronic copy can be an “article” in the same way as tangible articles for the purposes of secondary copyright infringement by importation into the UK, which would encompass an AI tool, Cerys Wyn Davies, a specialist at Pinsent Masons, says.
“This case is just the beginning of the complex picture in the battle between AI model developers and copyright holders,” Matthew Holman, a partner at Cripps, says.
The government told The Times that ministers are aware of the judgment and recognise its importance to those on both sides of the debate and are “committed to developing an approach that allows both to thrive”.
Last year the government published a controversial consultation proposing a change in the law to permit AI developers to train on proprietary material unless rights holders opted out. It received 11,500 responses that are being reviewed before final proposals are set out. As part of the Data (Use and Access) Act 2025, the government has also committed to publish an economic impact assessment, and report on the use of copyright works in the development of AI systems by March 18, 2026. A progress update is expected next week.
No comments:
Post a Comment