EU Rules AI-Generated Content Is Not Protected by Copyright: What This Means for Creators

The Copyright Dilemma of AI-Created Works

As generative AI experiences explosive growth, a fundamental legal question is becoming increasingly prominent: who owns the text, images, music, and even code created by AI? A recent topic that sparked heated discussion on Hacker News provided a clear answer—in the EU, AI-generated content is not protected by copyright. This position may seem straightforward, but it strikes at the core logic of the intellectual property system and casts a shadow over the entire AI content industry.

This topic received 33 upvotes and 34 comments on the Hacker News community, with the discussion intensity reflecting the widespread anxiety among tech professionals about AI copyright ownership. For creators, developers, and businesses that rely on AI tools for content production, this isn't just a legal concept—it's about the very foundation of their business models.

Generative AI: Understanding the Technical Background

To understand the copyright controversy, one must first understand how generative AI works. Generative AI refers to artificial intelligence systems capable of creating new content based on input prompts. Representative technologies include large language models based on the Transformer architecture (such as the GPT series, Claude), diffusion models (such as Stable Diffusion, DALL-E), and music generation models (such as Suno, Udio). These models learn statistical patterns and semantic relationships from massive datasets through pre-training, enabling them to generate content that is virtually indistinguishable from human creation on the surface. The launch of ChatGPT in late 2022 marked the explosion of generative AI into public awareness, after which AI content generation tools penetrated writing, design, programming, music, and nearly all creative industries at an astonishing pace.

It is precisely because the boundary between AI output and human creation is becoming increasingly blurred that the copyright ownership question has become so urgent and complex.

The Legal Foundation of the EU's Position

The "Human Author" Principle

The core of the EU copyright system is built on an ancient and robust principle: copyright protects the results of human intellectual creation. According to standards established by the Court of Justice of the European Union (CJEU) in multiple precedents, for a work to receive copyright protection, it must be a manifestation of the author's "own intellectual creation," reflecting the author's personality and demonstrating their free and creative choices.

The EU's copyright legal framework is primarily built on the 2001 Information Society Directive (InfoSoc Directive) and the 2019 Digital Single Market Copyright Directive (DSM Directive). The CJEU progressively established a unified standard for "originality" through a series of landmark cases, the most influential being the 2009 Infopaq case and the 2011 Painer case. In Infopaq, the CJEU first explicitly stated that originality requires a work to be the author's "own intellectual creation"; in Painer, the court further elaborated on the concept of creative choices, noting that even portrait photography can meet the originality threshold as long as the photographer made free creative choices regarding angle, lighting, composition, and other aspects. This set of standards applies to all EU member states, providing a solid legal basis for uniformly assessing the copyright status of AI-generated content.

In other words, copyright law protects not "the content itself" but the human creativity behind it. When a piece of text or an image is entirely generated by an algorithm based on prompts, it lacks genuine human creative choices and therefore cannot meet this "originality" threshold.

Comparison with the U.S. Copyright Office's Position

Notably, the EU's position is highly consistent with the U.S. Copyright Office's stance. The U.S. Copyright Office has repeatedly rejected copyright registration applications for purely AI-generated works, explicitly requiring that works must contain "human authorship" to receive protection.

The U.S. Copyright Office's attitude on AI copyright has been progressively clarified through a series of specific cases. In 2022, the office rejected Stephen Thaler's copyright registration application for a painting titled A Recent Entrance to Paradise independently generated by his AI system "Creativity Machine," citing the lack of human authorship. In 2023, in the case of Kristina Kashtanova's comic Zarya of the Dawn created using Midjourney, the Copyright Office made a more nuanced ruling: it acknowledged Kashtanova's copyright over the text and overall arrangement, but revoked copyright protection for the AI-generated images themselves. In March of the same year, the office issued formal policy guidance confirming that works containing AI-generated material require case-by-case review, with the core standard being whether a human exercised sufficient creative control over the final expression.

This means that across the two major transatlantic jurisdictions, attitudes toward AI content copyright are forming a kind of global consensus—machines cannot be authors.

What This Means for the AI Content Industry

Business Models Face Restructuring

For companies that treat AI-generated content as a core asset, this ruling presents substantive challenges. If the marketing copy, illustrations, or music you generate in bulk with AI cannot obtain copyright protection, then theoretically anyone can freely copy and use that content, and you cannot assert rights through legal means.

This directly impacts business models that attempt to build moats through "AI content factories." The industry must rethink: when content itself no longer possesses exclusivity, what should value be built upon? The answer may shift toward data, algorithms, brand, distribution channels, and deep human involvement.

"Human-AI Collaboration" Becomes the Key Variable

An important nuance is that the EU has not entirely denied the possibility of copyright for all works involving AI. The key lies in the degree of human involvement. If AI serves only as an assistive tool, and humans make substantive, creative choices during the creation process—such as deep editing, restructuring, or creatively integrating AI output—then the final work may still receive protection based on the human contribution.

What deserves deeper exploration here is the boundary between prompt engineering and creative choices. Prompt engineering refers to the technical practice of carefully designing input text to guide AI models toward specific outputs. The debate over whether prompts constitute "creative choices" is one of the most contentious topics in current copyright discussions. Proponents argue that complex prompt design involves deliberate consideration of style, composition, tone, and structure, similar to a director's guidance of actors. Opponents counter that prompts merely describe desired outcomes rather than controlling the specific form of expression—AI models may produce vastly different outputs from the same prompt, and users lack deterministic control over the final arrangement of specific pixels or words. Currently, the mainstream position of both the U.S. Copyright Office and the EU leans toward the latter view, holding that prompt input alone is insufficient to constitute creative choices in the copyright law sense.

This effectively points creators toward a path: AI should be the "brush," not the "painter." In the Hacker News discussion, many developers also noted that how to define "substantive human involvement" will become the focus of extensive future legal disputes.

Controversies and Unresolved Questions

The Dual Effect of the Copyright Vacuum

The Hacker News community discussion revealed the complexity of this issue. Some argued that it is reasonable for AI content to be unprotected, preventing the massive flood of machine-generated content from overwhelming and abusing the copyright system. Others worried that this "copyright vacuum" could trigger new chaos—when it is impossible to determine whether content is protected, corporate compliance risks and judgment costs will increase dramatically.

Additionally, there exists a thorny evidentiary problem: how do you prove whether a work is AI-generated or human-created? As AI output becomes increasingly indistinguishable from human work, the practical enforcement challenges may far exceed the legislation itself.

Currently, AI content detection technology faces serious reliability challenges. Mainstream detection methods include statistical feature-based analysis (such as text perplexity and burstiness metrics), watermarking technology (embedding detectable covert markers during the generation phase), and specially trained classifier models. However, OpenAI shut down its AI text detection tool in 2023 due to an excessively high false positive rate; research shows that AI content that has been slightly rewritten or human-edited is nearly impossible to reliably detect. In the image domain, while metadata standards like C2PA and implicit watermarking technologies have made progress, these markers can easily be removed or forged. This technological limitation makes "proving content origin" extremely difficult in practice, directly impacting the feasibility of copyright enforcement.

The Copyright Issue of Training Data Is Also Significant

Worth extending our thinking to is that the copyright issue in AI is "bidirectional." On one hand, there's the copyright ownership of AI output content (the core of this article's discussion); on the other hand, there's the question of whether AI input—i.e., training data—infringes on original creators' copyrights. These two issues together constitute a major challenge to the intellectual property system in the AI era. The EU's clear stance on the output side also foreshadows the cautious position it may adopt regarding input-side regulation.

In fact, the copyright issue of AI training data has already triggered a massive wave of litigation globally. In the United States, The New York Times v. OpenAI and Microsoft (filed in December 2023) is one of the most closely watched cases, with the newspaper alleging that its copyrighted articles were used to train GPT models and that the model can reproduce original text nearly verbatim. The Getty Images v. Stability AI case focuses on similar issues in the image domain. In the EU, Article 4 of the 2019 DSM Directive introduced a "text and data mining" (TDM) exception clause, permitting the use of copyrighted material for data mining and training under specific conditions, but rights holders can opt out through "machine-readable means." This institutional design attempts to find a balance between AI innovation and original creators' rights, but its actual effectiveness and enforcement mechanisms remain highly contentious. Japan has adopted a more permissive stance—its 2018 revised copyright law allows the use of copyrighted works for machine learning purposes, regardless of commercial or non-commercial use. Different countries' attitudes on the input side will profoundly affect the global competitive landscape of the AI industry.

The Law Is Catching Up with Technology

The EU's position that "AI-generated content is not protected by copyright" is essentially a natural extension of an ancient legal principle in the face of new technology, rather than a special restriction targeting AI. It sends a clear signal: for the foreseeable future, the law will continue to place human creativity at the center of the intellectual property system.

For AI practitioners, content creators, and businesses, this ruling is both a constraint and a reminder—truly valuable, protectable creation still cannot do without deep human involvement. In an era when AI can generate everything, human judgment, aesthetics, and creative choices may paradoxically become the scarcest assets—and the ones most favored by the law. As technology continues to evolve, the rules surrounding AI copyright will continue to be refined. This race between law and technology has only just begun.

Key Takeaways