🔗 Share this article

A AI firm based in the UK has won in a landmark judicial proceeding that examined the legality of machine learning systems using vast amounts of copyrighted data without permission.

Court Decision on Model Development and Intellectual Property

The AI company, whose leadership includes Oscar-winning director James Cameron, successfully defended against claims from Getty Images that it had violated the international image company's copyright.

Industry observers view this ruling as a blow to rights holders' exclusive right to benefit from their artistic output, with one senior lawyer warning that it indicates "the UK's current copyright regime is not adequately robust to protect its artists."

Evidence and Brand Concerns

Court documentation revealed that the agency's images were in fact employed to train the company's AI model, which enables individuals to generate images through written instructions. Nonetheless, Stability was also determined to have violated Getty's trademarks in certain cases.

The justice, Mrs Justice Joanna Smith, remarked that establishing where to strike the balance between the interests of the artistic industries and the AI industry was "of very real public importance."

Legal Complexities and Dismissed Claims

The photo agency had originally sued Stability AI for violation of its IP, alleging the technology company was "completely indifferent to what they fed into the training data" and had scraped and replicated countless of its images.

However, the agency had to withdraw its initial copyright case as there was insufficient evidence that the training took place within the United Kingdom. Alternatively, it continued with its suit arguing that the AI firm was still employing reproductions of its visual assets within its platform, which it called the "lifeblood" of its business.

Technical Complexity and Legal Reasoning

Highlighting the intricacy of artificial intelligence IP cases, the company fundamentally contended that the firm's visual creation system, known as Stable Diffusion, constituted an violating reproduction because its development would have represented copyright infringement had it been conducted in the United Kingdom.

The judge ruled: "A machine learning system such as Stable Diffusion which fails to retain or reproduce any protected works (and has not done so) is not an 'violating copy'." She elected not to rule on the misrepresentation allegation and found in favor of some of Getty's claims about brand infringement involving watermarks.

Industry Responses and Future Consequences

In a statement, Getty Images stated: "We remain deeply concerned that even financially capable companies such as Getty Images encounter significant challenges in safeguarding their artistic output given the lack of transparency requirements. Our company committed substantial sums of pounds to reach this point with only one provider that we need continue to address in a different venue."

"We encourage governments, including the United Kingdom, to implement stronger disclosure regulations, which are crucial to avoid costly legal battles and to enable creators to defend their interests."

Christian Dowell for the AI company commented: "We are satisfied with the court's ruling on the outstanding allegations in this case. Getty's decision to voluntarily dismiss the majority of its IP claims at the end of trial testimony left only a subset of claims before the judge, and this final decision ultimately addresses the copyright issues that were the core matter. Our company is grateful for the time and effort the court has put forth to settle the significant questions in this proceeding."

Wider Industry and Regulatory Context

This judgment emerges during an continuing debate over how the present government should regulate on the matter of copyright and artificial intelligence, with creators and writers including numerous well-known individuals lobbying for enhanced protection. Meanwhile, technology companies are advocating broad availability to copyrighted material to enable them to build the most advanced and effective AI creation platforms.

Authorities are currently consulting on copyright and AI and have stated: "Lack of clarity over how our intellectual property framework operates is impeding growth for our AI and creative industries. That cannot persist."

Legal specialists following the issue suggest that authorities are considering whether to implement a "text and data mining exemption" into British copyright law, which would allow copyrighted works to be utilized to train AI models in the UK unless the owner opts their content out of such training.