As generative AI continues to transform the fashion industry, complex copyright questions—particularly around the notion of authorship—are rapidly emerging. In this creative process, prompts play a pivotal role in shaping the human contribution, whether in visual design, music, audiovisual works, or even technical ideas and inventions. This shared reliance on human creative input suggests that a common legal framework may apply across these domains. In what follows, we explore the key legal boundaries and practical considerations that every designer and creator should understand.
Hey there! Have you seen the latest from New York Fashion Week? Brands like Collina Strada are making waves with unique prints and silhouettes created using generative AI. It’s official—AI is no longer just a novelty; it’s becoming a core part of the fashion industry. Students at FIT are using it to analyze trends, and one Hong Kong fashion show even produced over 80 AI-assisted garments.
AI is showing incredible potential in the early stages of brainstorming and concept development. But behind this dazzling technology lie some complex and sensitive legal questions. ‘Who actually owns the copyright to an AI-assisted design?’ and ‘What’s the risk of infringing on the countless existing designs the AI learned from?’ These are fundamental issues we’ll face across all creative fields. Today, we’re going to take a deep dive into this challenging but crucial topic, focusing on legal discussions in the EU and the UK. π
AI Meets Fashion Design: The Reality and Its Limits
The traditional fashion design process follows a long sequence: trend research, ideation, sketching, and prototyping. Generative AI has become a true ‘game-changer,’ especially in the initial brainstorming and concept development phases. It allows designers to dramatically save time and effort and find new inspiration when facing a creative block.
However, AI isn't a magic bullet. Studies have shown that while AI-generated sketches have high visual quality, they have significant limitations in originality and the ability to reflect a designer's detailed intent.
In one study, the ‘designer input and customization’ aspect of AI performance received the lowest scores. Yet, that’s precisely what designers value the most! This clearly shows that AI is unlikely to ever fully replace a designer's unique artistic sensibility.
The Core of Copyright: Was There a ‘Human Creative Choice’?
Let’s get to the most important question: who owns the copyright to an AI-assisted design? The fundamental principle of current law is ‘human-centric.’ This means works created solely by AI cannot receive copyright protection.
The issue arises when a human designer uses AI as a ‘tool.’ In the European Union (EU) and the UK, the standard of ‘Author’s Own Intellectual Creation (AOIC)’ is applied. The key here isn’t the aesthetic quality of the final product, but whether the designer made ‘free and creative choices’ that reflect their personality during the creation process.
A 3-Step Evaluation Process for Copyright
- Prompt Curation: The prompt a designer inputs into the AI is itself a result of creative choices. A detailed prompt like, “a red, midi-length, sleeveless A-line dress with ruffle details on the shoulders,” reflects the designer's original thought process.
- Output Selection and Modification: The designer’s ‘personal touch’ is added when they ‘select’ a specific design from numerous AI-generated outputs and ‘modify’ it to fit their vision.
- Final Completion: The designer’s personality and creativity are fully expressed when they add final details to the selected AI output and bring it to life as a physical garment.
Ultimately, if a human designer's creative intervention can be sufficiently proven throughout these stages, there's a possibility for the AI-assisted design to be protected by copyright.
The Complex Dilemma of Third-Party Infringement
While securing copyright for your own design is crucial, the risk of unknowingly infringing on someone else’s copyright is an even bigger concern. Generative AI learns through a process called Text-and-Data Mining (TDM), which scrapes vast amounts of data from the internet. This training data can include copyrighted content, and the process itself can be considered an act of ‘reproduction.’
Copyright infringement is typically determined by three factors: the act of reproduction, a causal link, and substantial similarity. The key is whether “the reproduction of a ‘substantial part’ that constitutes the ‘author's own intellectual creation’” of the original work has occurred.
When an AI model is excessively trained on certain data (overfitting), it may simply regurgitate something nearly identical to its training data instead of creating something new. For instance, there was a case where DALL-E produced strikingly similar images of a red dress—down to the length, neckline, and slit location—when given the same prompt. This is a serious red flag that could lead to unintentional copyright infringement.
To combat this risk, AI companies are implementing ‘AI output filtering technology’. Furthermore, with the EU’s new AI Act, which will require providers to disclose summaries of their training data, it will become easier to assess potential copyright infringement in the future.
Legal Defense Strategy: Claiming ‘Transformative Work’
Fortunately, there are ways to navigate these challenges. Even if an AI's initial output is similar to a copyrighted work, a designer can “create a new, non-infringing ‘derivative’ or ‘transformative’ work through sufficient modifications” that adds new meaning or message.
[Case Study] Spain’s ‘Vegap v Mango’ Case
In this case, a court ruled that transforming a copyrighted painting into a digital fashion item was not infringement. It determined the new piece was “‘transformative’ and non-infringing because it provided a new expression, meaning, message, or expanded utility.”
This is an important precedent, showing that designs based on AI-generated content can be considered independent works if they are sufficiently differentiated through the designer's creative intervention.
Therefore, designers using AI should be sure to follow these practical steps:
- Systematically Document Your Creative Process: It’s crucial that “the designer's personality and creative freedom are present and documented.” This record of your prompts and modifications is your strongest evidence in a legal dispute.
- Use Safe Tools: Remember that copyright exceptions may apply when using AI tools for “non-commercial or private research purposes,” and choose AI tools with clear licensing.
- Maximize Your Creative Contribution: Use AI outputs as a ‘first draft’ and focus on transforming them into something truly your own, “differentiated enough that the ‘substantial part’ of the original work is no longer recognizable in the new piece.”
AI Fashion Copyright: Key Takeaways
Frequently Asked Questions
Conclusion: The Future of Creativity and the Human Role
The copyright debate around AI-assisted fashion design reminds us of a crucial truth: no matter how advanced technology gets, the core of creation remains a uniquely human domain. AI is a powerful tool, but how that tool is used and in which direction it’s guided ultimately depends on the creative choices of a human designer.
In the new paradigm of human-AI collaboration, true value will come from human emotion, experience, and a unique perspective on the world. Amid the infinite possibilities that AI offers, it is still the human touch that elevates an output into a meaningful creation. If you have any more questions, feel free to leave a comment below!
This article is primarily based on the paper titled “Generative AI in fashion design creation: a copyright analysis of AI-assisted designs” (Lapatoura et al.), published in the Journal of Intellectual Property Law & Practice.
And this blog post is for general informational purposes only and cannot substitute for legal advice on specific matters. Please be sure to consult with a professional regarding individual legal issues.
