The UK creative industry’s attempt to prevent tech giants from exploiting copyrighted work to train AI models has been largely thrown out by the High Court, in a move which one legal expert insists leaves “the legal waters of copyright and AI training as murky as before”.
The case dates back to January 2023 when Getty Images started legal proceedings against Stability AI in both the UK and the US, filing a broad claim encompassing copyright, database right, trademark infringement and passing off allegations relating to the Stable Diffusion model, its training and development process and images generated by it.
During the trial, Getty Images withdrew its core claims for primary copyright and database infringement after acknowledging a lack of evidence that the training activities took place within the UK jurisdiction, a requirement under UK law.
While this meant the court could not rule on the core issue of whether using copyrighted material for AI training is lawful in the UK, it also exposed how hard it could be to stop overseas developers from doing the same.
The decision left two claims. The first was whether Stability’s making the model weights for certain versions of Stable Diffusion available for download was a secondary infringement of Getty’s copyright. The second was whether Stability had infringed either the Getty or iStock trademarks due to the presence of watermark-like features in certain Stable Diffusion outputs identified by Getty.
In response, Stability argued that the development and training of Stable Diffusion took place outside the UK. As Getty’s claim was limited to its UK copyright and database right, Stability argued that those rights had not been engaged by the process of making Stable Diffusion.
Stability also challenged Getty’s assertion that outputs from Stable Diffusion could infringe Getty’s copyright, database right and/or trade marks, or that Stability had committed any act of secondary copyright infringement by dealing with the models in the UK.
In the judgment, published earlier this week, the court dismissed the secondary copyright infringement, ruling that the Stable Diffusion model itself was not an “infringing copy” because it learns patterns from data rather than storing or reproducing the original copyrighted works.
Even so, Getty Images did succeed in the trademark infringement part on the claim. The court found that Stability AI had infringed Getty’s trademarks by producing AI-generated images that contained their watermarks. However, the judge noted these findings were “both historic and extremely limited in scope”, applying mainly to older models and isolated instances.
Meanwhile, the court did not provide a definitive ruling on the passing off claim.
James Clark, data protection, AI and digital regulation partner at law firm Spencer West LLP, told Decision Marketing: “At the heart of the judgment is the finding that the training of Stable Diffusion’s AI model using copyright work did not result in the production of an infringing copy of that work.
“It is the finding that the AI model did not store any copy of the protected works, and the model itself was not an infringing copy of such work, that will cause concern for the creative industry, while giving encouragement to AI developers.”
Clark added that the judgment highlights the problem that the creative industry has in bringing a successful copyright infringement claim in relation to the training of large language models. “During the training process, the model is not making a copy of the work used to train it, and it does not reproduce that work when prompted for an output by its user. Rather, the model ‘learns’ from the work, in a similar way to the way a human might.”
Meanwhile, Nathan Smith, IP partner at Katten Muchin Rosenman LLP, added: “While Getty succeeded in part on limited trademark infringement claims relating to the unauthorised outputs of ‘iStock’ and ‘Getty Images’ watermarks by earlier models, these findings were based on specific examples and offer minimal practical impact.
“By dismissing Getty’s secondary infringement claims, the judgment appears to present a win for the AI community, but arguably leaves the legal waters of copyright and AI training as murky as before.”
The US case continues.
The minister of state for media, tourism and creative industries Chris Bryant has long insisted that Government will consider changes to the law to protect the UK creative industry from such issues but the result of its copyright and AI consultation – which closed in February this year – have yet to be released.