🔗 Share this article

A AI company headquartered in London has won in a significant high court proceeding that examined the legality of machine learning systems using extensive quantities of protected data without authorization.

Court Decision on AI Training and Intellectual Property

The AI company, whose leadership includes Academy Award-winning filmmaker James Cameron, effectively defended against claims from the photo agency that it had violated the international photo company's copyright.

Legal experts view this decision as a setback to copyright owners' sole right to benefit from their artistic work, with a prominent attorney warning that it indicates "Britain's current IP system is not adequately robust to protect its artists."

Evidence and Brand Concerns

Court evidence revealed that the agency's images were indeed employed to train Stability's system, which enables individuals to create images through text prompts. Nonetheless, Stability was also determined to have infringed the agency's brand marks in some cases.

The presiding judge, Mrs Justice Joanna Smith, remarked that establishing where to strike the equilibrium between the concerns of the creative industries and the artificial intelligence industry was "of very real societal importance."

Legal Complexities and Dismissed Allegations

Getty Images had originally filed suit against the AI company for infringement of its IP, alleging the technology company was "entirely indifferent to what they fed into the training data" and had collected and replicated millions of its images.

However, the agency had to withdraw its original IP claim as there was insufficient evidence that the development occurred within the UK. Alternatively, it continued with its suit claiming that Stability was still using copies of its image content within its systems, which it called the "lifeblood" of its operations.

Technical Complexity and Judicial Analysis

Demonstrating the complexity of AI copyright cases, the company fundamentally argued that the firm's image-generation system, known as Stable Diffusion, constituted an infringing copy because its creation would have constituted copyright violation had it been conducted in the United Kingdom.

The judge ruled: "An AI model such as Stable Diffusion which does not store or replicate any protected works (and has never done) is not an 'infringing reproduction'." She declined to make a determination on the passing off claim and found in support of certain of the agency's claims about trademark violation involving digital marks.

Industry Reactions and Future Consequences

In a statement, Getty Images said: "We continue to be deeply concerned that even well-resourced companies such as Getty Images encounter significant challenges in protecting their creative works given the absence of transparency standards. We invested millions of pounds to reach this point with only one provider that we need continue to address in a different forum."

"We urge authorities, including the UK, to implement more robust transparency rules, which are essential to prevent expensive court proceedings and to enable creators to protect their rights."

Christian Dowell for Stability AI said: "We are pleased with the judicial decision on the remaining claims in this case. The agency's decision to voluntarily dismiss the majority of its IP cases at the end of court testimony resulted in a subset of claims before the judge, and this final ruling eventually addresses the IP issues that were the core matter. We are grateful for the time and consideration the court has put forth to settle the important questions in this case."

Broader Industry and Regulatory Background

The judgment emerges during an ongoing discussion over how the current administration should regulate on the issue of copyright and AI, with artists and writers including several well-known individuals lobbying for greater protection. Meanwhile, tech firms are advocating broad availability to protected content to allow them to build the most powerful and efficient generative AI systems.

The government are presently seeking input on IP and AI and have stated: "Uncertainty over how our intellectual property framework operates is holding back development for our artificial intelligence and creative industries. That must not continue."

Industry experts monitoring the situation indicate that regulators are examining whether to introduce a "content analysis exception" into British copyright legislation, which would permit protected works to be utilized to train AI models in the UK unless the rights holder opts their works out of such training.