Courts Face AI Copyright Crossroads in 2025

Courts across the United States and Europe now face a defining legal question of the AI era: does art created by artificial intelligence deserve copyright protection? With the U.S. Copyright Office, federal judges, and international tribunals all grappling with conflicting rulings and mounting pressure from both creators and AI companies, 2025 is shaping up as the year the legal world can no longer punt on this issue.

The stakes are enormous. The global generative AI market is projected to reach $137 billion by 2030, according to Grand View Research, and a significant share of that value hinges on who owns what AI systems produce. From Midjourney illustrations to DALL-E marketing assets to Stable Diffusion concept art, millions of images are generated daily — and their legal status remains in limbo.

Key Takeaways at a Glance

• The U.S. Copyright Office has repeatedly denied copyright registration for purely AI-generated works, citing the absence of human authorship
• Artist Jason Allen's Midjourney-created piece 'Théâtre D'opéra Spatial' was denied copyright in September 2023, and his appeal remains a bellwether case
• The European Union's AI Act, which took effect in August 2024, does not directly address copyright ownership of AI outputs
• Courts in China have taken a more permissive approach, granting copyright to AI-generated images in a landmark Beijing ruling
• Estimates suggest over 15 billion AI images have been generated since the launch of DALL-E 2 in 2022
• Major companies like Adobe, Getty Images, and Shutterstock have adopted conflicting policies on AI-generated content licensing

The Human Authorship Requirement Creates a Legal Wall

At the heart of the debate sits a deceptively simple principle: copyright law in the United States requires human authorship. The U.S. Copyright Office has held this position consistently, reaffirming it in its March 2023 guidance on AI-generated works. Under current interpretation, a person must exercise creative control over the expressive elements of a work to claim ownership.

This creates an immediate problem. When a user types a prompt into Midjourney or DALL-E 3, how much creative control are they actually exercising? The Copyright Office has argued that typing 'a sunset over a cyberpunk city in the style of Monet' is more akin to giving instructions to a commissioned artist than creating the art yourself. The AI, not the human, makes the countless micro-decisions about composition, color, lighting, and texture.

Compare this to photography, which faced similar skepticism in the 19th century. Courts eventually ruled that photographers exercise sufficient creative judgment — framing, timing, exposure — to qualify as authors. Proponents of AI copyright argue that prompt engineering involves analogous creative decisions, especially when users iterate through dozens or hundreds of variations to achieve a specific vision.

The Jason Allen Case Sets a Critical Precedent

Jason Allen became the unlikely face of this debate when his Midjourney-generated artwork won first place at the 2022 Colorado State Fair fine arts competition. The backlash was immediate, but the legal saga that followed proved even more consequential. Allen applied for copyright registration and was denied by the Copyright Office in September 2023.

Allen's appeal argues that he spent over 80 hours refining his prompts, curating outputs, and making post-processing edits in Adobe Photoshop. His legal team contends that this level of involvement constitutes sufficient human authorship. The case could establish whether the degree of human input — not just its presence — determines copyrightability.

If Allen prevails, it would open the door to a spectrum-based approach to AI copyright, where works involving significant human curation and editing receive protection while fully automated outputs do not. If he loses, it reinforces a bright-line rule that could leave billions of dollars worth of AI-generated content in the public domain.

China and the EU Chart Different Courses

The international landscape adds complexity. In November 2023, the Beijing Internet Court ruled that an AI-generated image could receive copyright protection, provided the user demonstrated 'intellectual investment' in the creation process. The court emphasized the plaintiff's selection of prompts, parameters, and the final image from multiple outputs as evidence of originality.

This stands in stark contrast to the U.S. approach and could create significant jurisdictional conflicts as AI-generated content flows across borders. A piece of art copyrighted in China might have zero protection in the United States, creating enforcement nightmares for global platforms.

The European Union has taken yet another path. The EU's AI Act focuses primarily on safety, transparency, and risk classification rather than intellectual property ownership. However, the EU Copyright Directive requires that works reflect the 'author's own intellectual creation,' a standard that some legal scholars believe could accommodate AI-assisted works more readily than U.S. law.

Key international positions include:

• United States: No copyright for works without human authorship; pending appeals could modify this
• China: Copyright possible if user demonstrates sufficient intellectual investment
• European Union: Unclear under current directives; member states may diverge
• United Kingdom: Has a specific provision (Section 9(3) of the CDPA 1988) that grants copyright to the person who makes 'arrangements necessary' for computer-generated works
• Japan: Generally permits AI-generated works to enter the public domain but is reviewing its position

The Business Implications Run Into the Billions

For companies building products around generative AI, the copyright question is not academic — it is existential. Adobe has invested heavily in its Firefly model, which is trained exclusively on licensed and public domain content, partly to offer customers legal indemnification. If AI-generated outputs cannot be copyrighted, that indemnification promise loses much of its value.

Shutterstock and Getty Images have both launched AI generation tools integrated into their platforms, with licensing frameworks that assume some form of ownership over the outputs. Getty has gone further, offering legal protection to enterprise customers who use its AI tools. But these guarantees rest on untested legal ground.

The advertising industry alone spends an estimated $7.5 billion annually on stock imagery and custom visual content, according to IBISWorld. If AI-generated visuals cannot be copyrighted, companies cannot prevent competitors from using identical or near-identical assets. This fundamentally undermines the value proposition of custom AI-generated marketing materials.

Startups are also affected. Firms like Jasper, Canva, and Runway have built multi-billion-dollar valuations on the assumption that their AI-generated outputs have commercial value tied to exclusivity. Without copyright protection, that exclusivity evaporates.

Artists and Creators Push Back From Both Sides

The creative community is deeply divided. Traditional artists — many of whom are already pursuing class-action lawsuits against companies like Stability AI, Midjourney, and DeviantArt — generally oppose copyright protection for AI outputs. Their argument is twofold: AI models were trained on copyrighted human art without consent, and granting copyright to AI outputs would flood the market with cheap substitutes that devalue human creativity.

On the other side, a growing cohort of AI-native artists argues that their work involves genuine creative expression. These creators often spend hours or days crafting intricate prompt sequences, combining multiple AI tools, and performing extensive manual editing. They view the denial of copyright as a failure to recognize a legitimate new art form.

The tension mirrors historical debates over the copyrightability of photographs, recorded music, and digital art. In each case, the legal system eventually adapted — but the transition period created uncertainty that lasted years or even decades.

What This Means for Developers and Businesses

Practical implications are already reshaping business strategy across the tech industry. Companies and individuals working with AI-generated content should consider several factors:

• Document human involvement: Keep detailed records of prompt iterations, manual edits, and creative decisions to strengthen potential copyright claims
• Layer human creativity: Use AI outputs as a starting point, then add substantial human modifications to increase the likelihood of protection
• Monitor jurisdictional differences: Content distributed globally may have different legal status in different countries
• Review contracts carefully: Licensing agreements for AI tools often contain clauses about output ownership that may conflict with copyright law
• Consider trade secret protection: Where copyright is unavailable, other intellectual property mechanisms may offer partial protection

Looking Ahead: The Clock Is Ticking

Several developments in the next 12 to 18 months could reshape this landscape dramatically. The Jason Allen appeal is expected to produce a ruling that will either reinforce or soften the Copyright Office's hard line. Multiple federal lawsuits involving AI training data — including cases against OpenAI, Meta, and Stability AI — could produce rulings with implications for output ownership.

The U.S. Congress has also begun to engage. The Senate Judiciary Committee held hearings on AI and copyright throughout 2023 and 2024, and legislative proposals are circulating that would create a new category of protection for AI-assisted works — potentially shorter in duration or narrower in scope than traditional copyright.

The most likely outcome, according to legal experts at firms like Morrison Foerster and Latham & Watkins, is a tiered system. Purely AI-generated works with minimal human input would remain unprotectable. Works involving substantial human curation, editing, and creative direction would receive some form of copyright. And works where AI serves merely as a tool — similar to how Photoshop or a digital camera functions — would receive full protection.

Whatever courts decide, the ruling will ripple far beyond the art world. It will set precedents for AI-generated code, music, writing, and video — every domain where generative AI is rapidly replacing or augmenting human labor. The $137 billion question is not whether the law will adapt, but whether it will adapt fast enough to prevent a prolonged period of commercial chaos.