A UK High Court recently ruled that Stability AI's Stable Diffusion image generator did not infringe Getty Images' copyright, even though the AI model was trained on millions of images. The decision hinges on a technical distinction in copyright law: while Stable Diffusion learned patterns from Getty's photos, it doesn't store or reproduce those images in its final model. Yet the ruling has exposed a fundamental tension between what the law says and what most people instinctively feel about AI using human creative work.

What Did the Getty Images v Stability AI Case Actually Decide?

Getty Images, the massive stock photo company, sued Stability AI, alleging that the company had used Getty's images without permission to train Stable Diffusion, the popular text-to-image generator that produces "photo-realistic images" from written prompts. Getty brought two types of copyright claims: primary infringement (direct copying) and secondary infringement (distributing products made from copied work).

The primary infringement claim was dropped because Getty couldn't prove that Stability AI had trained its model using UK-based images. But the secondary infringement claim went to trial. Getty argued that Stable Diffusion itself was an "infringing copy" under UK copyright law, the Copyright, Designs and Patents Act 1988 (CDPA). The High Court disagreed. Under the CDPA, an infringing copy must actually reproduce the original work. Stable Diffusion's model, the court found, doesn't store or reproduce Getty's images; it learns statistical patterns from them. Learning patterns is not the same as copying, legally speaking.

Why Does the Legal Ruling Feel Wrong to So Many People?

The court's decision is legally sound, but it clashes with a widespread intuition: that something is being "taken" from artists when AI systems train on their work. This feeling is not fringe. Roughly 70% of US adults believe that artists should be compensated when generative AI uses their work to produce images, according to polling cited in the case. The sense of loss is real for many creators. One artist, Greg Rutkowski, discovered that his work had been processed by AI models more than 400,000 times in under five years, without compensation or consent.

The tension stems from two core concepts in copyright law: originality and authorship. Copyright was designed to protect human creators. But generative AI breaks the traditional authorship model. When a human artist learns from another artist, both are recognized as creators with rights and responsibilities. When an AI model trains on millions of images, it's mapping statistical relationships between pixels and words, not exercising imagination in the way humans do.

How Does Copyright Law Define Authorship in the Age of AI?

The CDPA defines an "author" as "the person who creates." For computer-generated work, it specifies that the author is "the person by whom the arrangements necessary for the creation of the work are undertaken." This definition was written with late-20th-century automated systems in mind, like weather charts or formatted reports, where human labor could still be traced through the finished product.

Generative AI shatters that framework. When Stable Diffusion generates an image, multiple unrelated parties are involved: the developers who built the model, the company that fine-tuned it, and potentially millions of artists whose work trained the system. The law has no clear answer for who the "author" is or whose work is being protected.

The court acknowledged this discomfort. Even the UK government, when the humanoid robot artist Ai-Da unveiled an AI-painted portrait of King Charles III, conceded that the work raised "timely questions about the nature of creativity, authorship, and the future of art in the digital age".

Steps to Understanding the Legal and Ethical Landscape Around AI Training

• Understand the Copyright Distinction: Copyright law protects expression, not ideas or techniques. An art student who internalizes a thousand paintings and paints in their style infringes nothing, because style is an "idea." Generative AI models work similarly, learning patterns rather than copying specific works, which is why courts have ruled they don't infringe copyright in the traditional sense.
• Recognize the Authorship Gap: Current copyright law assumes a single human author or a clear chain of responsibility. Generative AI involves multiple parties with no clear relationship to one another, making it impossible to apply traditional authorship rules. This gap is not a legal oversight but a fundamental mismatch between the technology and the law's assumptions.
• Consider the Compensation Question: While the Getty ruling found no copyright infringement, it did not address whether artists should be compensated for their work being used in training. That is a policy question, not a legal one, and it remains unresolved in most jurisdictions.
• Examine the Imagination Question: Humans have imagination defined as the power to form "mental images of something not present to the senses or never before wholly perceived in reality." Generative AI produces no new style or idea drawn from experience or emotion. This philosophical difference may matter more than the legal technicality.

What Happens Next for AI Developers and Artists?

The Getty ruling does not end the debate. It clarifies that under current UK copyright law, training AI models on copyrighted images is not infringement, provided the model doesn't store or reproduce those images. But this legal clarity has not satisfied the broader concern. The discovery of Project Panama, an operation by Anthropic where the company purchased, scanned, and destroyed millions of books to train its AI chatbot Claude, sparked outrage among authors and publishers, showing that legal permission and public acceptance are not the same thing.

The real question the Getty ruling leaves unanswered is whether copyright law and human instinct are still describing the same thing. The law protects expression and authorship in ways designed for human creators. Generative AI operates in a space where those protections don't clearly apply. Until lawmakers, technologists, and artists find a new framework that addresses both the legal reality and the ethical intuition, this tension will persist.