When AI Copies Your Style: IP in the AI Era

A fantasy artist discovers his name has been used in AI prompts over 400,000 times in under a year, generating images that mimic his style so convincingly that he can no longer distinguish his own work from the imitations online. Tim Burton describes the digital reproduction of his style as "like a robot taking your humanity, your soul."

These are not hypothetical scenarios. They are documented cases that illustrate a gap in intellectual property law that most legal practitioners have not yet fully grappled with.

The problem is not reproduction — it is substitution

Traditional IP frameworks were designed to protect against unauthorized reproduction and distribution of works. The printing press, lithography, digital copying — each technological shift prompted legal adaptation around the same core question: who controls the copies?

Generative AI breaks this pattern. The threat is not that someone copies a specific painting. The threat is that someone can now produce new paintings that are stylistically indistinguishable from an artist's body of work — at scale, instantly, and at near-zero cost. The economic harm comes from substitution of the creator, not reproduction of the creation.

For visual artists who depend on commissions rather than licensing reproduction rights, this distinction is existential. If a client can prompt an AI to generate "illustration in the style of [Artist X]" and get commercially acceptable results, the economic incentive to commission Artist X disappears — regardless of whether any specific copyrighted work was reproduced.

Why copyright law struggles with style

Copyright protects expressions of ideas, not ideas themselves. Artistic style sits uncomfortably between these categories. A style is not a single work — it is a pattern of creative choices that emerges across many works. It encompasses brushwork, colour palette, compositional tendencies, subject matter preferences, and what philosophers of art call "dawning of an aspect" — the holistic impression that allows an observer to attribute an unsigned work to its creator.

The difficulty is well illustrated by Monet and Renoir. Both impressionists sometimes painted the same subjects side by side. An expert can distinguish Monet's loose brushwork and play of light from Renoir's warmer tones and blended technique. But if copyright were to protect these stylistic elements, where would the line fall? Should Renoir be prohibited from painting water lilies? Must he confine himself to warm tones to avoid infringing Monet's "style"?

Human artistic development has always involved imitation. Artists learn general style conventions and adapt elements from predecessors to develop their own voice. This is not a bug — it is how creativity works. Granting monopoly rights over artistic style would risk freezing the very dynamic that produces cultural richness.

The scenes a faire doctrine in US copyright law captures this intuition: elements that are standard or necessary within a genre cannot be monopolized. But AI-generated style reproduction operates in a grey zone — not copying specific scenes, but capturing the holistic pattern that makes an artist recognizable.

Moral rights: the stronger foundation

If copyright's economic rights framework is a poor fit for style protection, moral rights offer a more promising avenue. Article 6 bis of the Berne Convention establishes two key rights: the right to claim authorship, and the right to object to derogatory treatment of one's work.

For digital style reproduction, these rights are directly relevant in several ways.

False attribution. When an AI generates images "in the style of [Artist X]" and these circulate commercially — sometimes explicitly referencing the artist's name, sometimes implicitly — the market is flooded with works the artist never created. Some jurisdictions grant a specific moral right against false attribution, allowing individuals to object to their names being associated with works they did not create. This right protects against reputational harm even when no specific copyrighted work has been reproduced.

Derogatory treatment and brand dilution. AI-generated works that mimic an artist's style may be of inferior quality, may express messages the artist would reject, or may be used in contexts the artist would find objectionable. The CJEU addressed a related concern in Deckmyn, holding that even the parody exception must respect fundamental rights — including not being associated with discriminatory messages. For an artist whose name is used to prompt AI-generated content, the potential for reputational damage is substantial and largely uncontrolled.

The text and data mining question

The EU's CDSM Directive Article 4 provides a copyright exception for text and data mining with a rights reservation ("opt-out") mechanism. In theory, artists can reserve their rights to prevent their works from being used as AI training data. In practice, this approach has significant limitations.

The opt-out mechanism is vaguely defined and difficult to enforce. Artists often do not know whether their works are included in training datasets. Even when opt-out is technically available, the power asymmetry between individual artists and large AI companies makes meaningful enforcement impractical.

More fundamentally, the training data question may become irrelevant for style reproduction. Recent few-shot learning techniques can capture an artist's style from just a handful of images. Midjourney's "style reference" feature lets users upload images and instruct the model to replicate their style — bypassing any training data opt-out entirely. Voice cloning models can replicate an individual voice from seconds of audio. The technology is moving faster than the regulatory framework.

Indirect remedies and unfair competition

Beyond copyright, several legal doctrines may offer partial protection. Trademark law protects famous marks against uses that diminish their uniqueness. The CJEU in L'Oreal and Others addressed parasitism — taking unfair advantage of a mark's "power of attraction, its reputation, and its prestige" by riding on its coattails.

For artists whose names function as brands, unfair competition concepts including slavish imitation and unfair free-riding may apply — particularly when a commercial entity deliberately invokes the artist's name to benefit from their reputation. Some scholars have also proposed recognizing an individual artistic style as personal data under the GDPR, or applying rights of publicity.

These are indirect remedies with high legal uncertainty. They work best when there is a clear commercial use of the artist's name or brand. They become much harder to apply when the stylistic similarity exists without explicit name reference — bringing us back to the fundamental challenge of defining artistic style with legal precision.

In Switzerland specifically, the Federal Act against Unfair Competition (UWG) provides additional angles. Article 3(1)(e) UWG addresses the risk of confusion through imitation of third-party products or services, and the prohibition against exploitation of another's reputation could be extended to cover systematic commercial use of an artist's stylistic identity. Swiss personality rights under Article 28 of the Civil Code add another layer — the right to protection of one's identity, which could encompass a distinctive artistic style developed over a career, is well-established in Swiss jurisprudence. These domestic instruments deserve more attention from practitioners advising creative clients in the Swiss market.

What practitioners should be doing now

The law will evolve, but creative clients need protection today. Here is where advisory work should focus.

Contract design. Artists entering commercial relationships should negotiate explicit provisions covering digital style reproduction. Performers' unions have already confronted similar issues around digital likeness rights, and the lessons are directly applicable. Contract language must be clear and accessible — not buried in ambiguous clauses that artists sign without understanding.

Proactive rights management. Help clients establish clear documentation of their stylistic identity, maintain attribution records, and implement opt-out mechanisms where available. This creates an evidentiary foundation for future claims.

Monitor for misuse. Tools exist to track when an artist's name appears in AI-generated content databases and marketplaces. Early detection of brand dilution enables faster response.

Moral rights enforcement. In jurisdictions with strong moral rights provisions, false attribution and integrity claims can be pursued now — they do not require waiting for new legislation.

The fashion and design angle

It is worth noting that the style reproduction problem extends well beyond fine art. In fashion and product design, distinctive visual identities are core business assets. The Swiss Federal Design Act (Designgesetz) protects the appearance of industrially produced or crafted products — but only specific designs, not a broader stylistic language. When AI enables anyone to generate products "in the aesthetic of [Brand X]," the gap between design registration and style protection becomes commercially significant.

The fashion industry has long grappled with the tension between inspiration and copying. Fast fashion already demonstrated that stylistic elements — silhouettes, colour combinations, material choices — are difficult to protect under existing IP frameworks. AI accelerates this dynamic by orders of magnitude. A design house that took weeks to interpret a runway collection can now be undercut by AI systems that generate derivative designs in minutes.

For practitioners advising fashion and design clients, the contractual and proactive strategies outlined above are even more urgent. Unlike fine artists, commercial designers often operate within employment or commission relationships where IP ownership clauses are negotiated. Ensuring these clauses explicitly address AI-generated style reproduction is a concrete, actionable step.

The road ahead

The chapter of IP law on AI and creative style is being written now. Current frameworks were not designed for a world where the creation of works — not just their reproduction — can be automated. The gap between what the law protects and what artists need is real and growing.

The most likely trajectory involves adaptation rather than revolution: expanding moral rights frameworks, clarifying TDM regulations, developing new remuneration models for name-based style reproduction, and potentially creating sui generis rights for distinctive artistic identities. The EU AI Act and ongoing national implementations will add regulatory layers that intersect with IP protection in ways that are not yet fully mapped.

For legal practitioners in Switzerland and the DACH region, this is not a niche topic. Creative industries are significant economic sectors, and the clients who need sophisticated IP counsel on these questions are already walking through the door. The practitioners who understand both the technology and the evolving legal landscape will be the ones who provide real value.