An artificial intelligence company based in London has achieved a significant victory in a High Court case that scrutinized the legality of an AI model using extensive copyrighted data without authorization.

Stability AI, led by Oscar-winning Avatar director James Cameron, successfully defended itself against allegations from Getty Images, claiming that it infringed on the international photography agency’s copyright.

This ruling is seen as a setback for copyright holders’ exclusive rights to benefit from their creations. Rebecca Newman, a legal director at Addleshaw Goddard, cautioned that it suggests “the UK derivative copyright system is inadequate to protect creators”.

There was evidence indicating that Getty Images were utilized in training Stability’s model, which enables users to generate images via text prompts. In certain instances, Stability was also found to violate Getty’s trademarks.

Judge Joanna Smith remarked that determining the balance between the interests of the creative industries and AI sectors holds “real social significance.” However, she could only address relatively limited claims as Getty had to withdraw parts of its case during the trial this summer.

Getty Images initiated legal action against Stability AI for violations of its intellectual property rights, claiming the AI company scraped and replicated millions of images with “complete indifference to the content of the training data.”

This ruling comes amid ongoing debates about how the Labour government should legislate on copyright and AI matters, with artists and authors like Elton John, Kate Bush, Dua Lipa, and Kazuo Ishiguro advocating for protections. In contrast, tech firms are seeking broader access to copyrighted material to develop more powerful generative AI systems.

The government is conducting a consultation regarding copyright and AI, stating: “The uncertainty surrounding the copyright framework is hindering the growth of both the AI and creative sectors. This situation must not persist.”

Lawyers at Mishcon de Reya, pursuing this matter, are contemplating introducing a “text and data mining exception” to the UK copyright law, which would enable copyrighted works to be utilized for training AI models unless rights holders opt-out.

Due to a lack of evidence indicating that the training took place in the UK, Getty was compelled to retract its original copyright claim. Nevertheless, the company proceeded with its lawsuit, asserting that Stability continues to use copies of visual assets, which it describes as the “lifeblood” of its business. The lawsuit alleges trademark infringement and “spoofing,” as some generated images bore Getty’s watermark.

Highlighting the complexities of AI copyright litigation, the group essentially argued that Stability’s image generation model, known as Stable Diffusion, constitutes an infringing copy, as its creation would represent copyright infringement if produced in the UK.

The judge determined that “AI models like Stable Diffusion that do not (and never have) stored or reproduced copyrighted works are not ‘infringing copies.'” She declined to adjudicate on the misrepresentation claims but ruled in favor of some of Getty’s trademark infringement claims regarding the watermark.

In a statement, Getty Images remarked: “We are profoundly worried that even well-resourced organizations like Getty Images face considerable challenges in safeguarding creative works due to the absence of transparency requirements. We have invested millions with one provider alone, but we must continue our pursuit elsewhere.”

“We urge governments, including the UK, to establish more robust transparency regulations. This is crucial to avoid expensive legal disputes and ensure creators can uphold their rights.”

Stability AI’s General Counsel, Christian Dowell, stated, “We are pleased with the court’s ruling on the remaining claims in this case. Although Getty’s decision to voluntarily withdraw most of the copyright claims at the trial’s conclusion left the court with only a fraction of the claims, this final decision addresses the core copyright issues. We appreciate the time and effort the court has dedicated to resolving the significant matters in this case.”