🔗 Share this article

A AI company based in London has won in a significant high court proceeding that examined the lawfulness of machine learning systems using vast amounts of copyrighted material without permission.

Judicial Ruling on AI Training and Intellectual Property

Stability AI, whose directors includes Academy Award-winning filmmaker James Cameron, successfully resisted allegations from Getty Images that it had infringed the global image company's intellectual property rights.

Legal experts view this decision as a blow to copyright owners' exclusive right to profit from their creative output, with one prominent attorney warning that it demonstrates "the UK's secondary IP system is not adequately robust to safeguard its artists."

Evidence and Brand Concerns

Judicial documentation revealed that Getty's photographs were in fact used to train the company's system, which enables users to generate visual content through text instructions. Nonetheless, the AI firm was also found to have infringed Getty's trademarks in certain instances.

The presiding justice, Mrs Justice Joanna Smith, stated that establishing where to strike the balance between the interests of the creative sectors and the artificial intelligence sector was "of very real societal importance."

Legal Challenges and Withdrawn Claims

The photo agency had originally filed suit against the AI company for infringement of its IP, alleging the AI firm was "entirely unconcerned to what they fed into the development material" and had scraped and replicated millions of its photographs.

However, the agency had to withdraw its initial IP claim as there was insufficient evidence that the training occurred within the UK. Instead, it proceeded with its suit arguing that Stability was still employing reproductions of its image content within its systems, which it called the "lifeblood" of its business.

Technical Intricacy and Judicial Analysis

Highlighting the intricacy of artificial intelligence IP cases, the agency fundamentally argued that Stability's visual creation system, called Stable Diffusion, amounted to an violating reproduction because its creation would have constituted copyright violation had it been conducted in the United Kingdom.

Mrs Justice Smith determined: "A machine learning system such as Stable Diffusion which fails to retain or replicate any protected material (and has never done) is not an 'violating reproduction'." She declined to make a determination on the misrepresentation claim and found in support of some of the agency's arguments about brand violation related to watermarks.

Sector Responses and Ongoing Implications

Through a official comment, the photo agency said: "We continue to be profoundly worried that even well-resourced companies such as our company encounter significant challenges in safeguarding their artistic output given the lack of disclosure standards. Our company committed millions of pounds to achieve this stage with only a single company that we must proceed to address in another forum."

"We urge authorities, including the United Kingdom, to establish more robust transparency rules, which are crucial to avoid costly legal battles and to allow creators to defend their rights."

Christian Dowell for Stability AI commented: "We are pleased with the judicial ruling on the outstanding allegations in this case. The agency's decision to voluntarily dismiss the majority of its copyright cases at the conclusion of trial proceedings resulted in a subset of allegations before the judge, and this concluding ruling eventually addresses the copyright concerns that were the central issue. We are thankful for the attention and effort the court has put forth to settle the significant questions in this proceeding."

Broader Industry and Regulatory Context

The ruling comes during an continuing debate over how the present government should legislate on the matter of intellectual property and AI, with creators and writers including several prominent individuals advocating for enhanced protection. At the same time, technology companies are advocating broad availability to protected content to enable them to build the most powerful and effective generative AI platforms.

Authorities are currently seeking input on IP and AI and have declared: "Lack of clarity over how our intellectual property system functions is impeding development for our artificial intelligence and creative sectors. That must not persist."

Industry experts monitoring the issue suggest that regulators are examining whether to implement a "text and data mining exception" into British copyright legislation, which would permit protected material to be used to train machine learning systems in the United Kingdom unless the owner opts their works out of such development.