Artificial Intelligence | Intellectual Property Law | Technology

Introduction: The New Creators Have No Pulse

In the past three years, artificial intelligence has evolved from an experimental novelty to a prolific creator—generating poems, symphonies, digital paintings, codebases, and video games. Whether it’s a TikTok filter that renders dreamlike animations or a large language model drafting screenplays, today’s most powerful creators aren’t human. And that raises one fundamental legal question: who owns what the algorithm creates?

The answer lies at the murky crossroads of intellectual property law and emerging AI regulation—a battleground where copyright, authorship, and machine autonomy collide.

I. The Legal Vacuum: AI Outputs and Copyright Protection

Under U.S. law, copyright is reserved for “original works of authorship” fixed in a tangible medium (17 U.S.C. § 102). Critically, authorship must be human—a position reaffirmed repeatedly by courts and agencies.

Key Precedent: Thaler v. Perlmutter (2023)

In a case involving AI-generated artwork submitted without a human author, the D.C. District Court sided with the U.S. Copyright Office, ruling that creations generated solely by AI are not eligible for copyright protection. The court concluded:

“Human authorship is a bedrock requirement of copyright.”

This decision aligns with the Copyright Office’s 2023 guidance: AI-generated works may receive protection only if a human has exercised meaningful creative control—not merely clicked “generate.”

II. Who Owns AI-Generated Work? It Depends

Because purely AI-generated works fall outside copyright protection (at least under current law), questions of ownership shift to contractual arrangements and software licenses. Key ownership structures now include:

1. Human + AI Collaboration

If a human prompts an AI tool (e.g., by selecting styles, editing outputs, combining content), they may claim authorship of the human-directed portions. Courts may recognize limited copyrights based on the degree of human input and judgment.

2. Work Made for Hire (WMFH)

Companies may own AI-generated assets through WMFH agreements with employees or contractors. But ambiguity remains: is the AI the “author,” or is it a tool used by the human creator?

3. License-by-Use

Platforms like OpenAI, Adobe Firefly, and Midjourney assign broad usage rights to users, granting nonexclusive licenses to generated content—though these may not convey full copyright title.

Practice Tip: Ensure your contracts address IP ownership of AI-assisted work, particularly in branding, advertising, and software development.

III. What About Infringement? Tracing AI’s Data Trail

Ownership is just half the issue. The other is liability: can AI-created content infringe on someone else’s copyright? The answer is yes—especially when:

• AI is trained on copyrighted materials without consent
• Outputs closely replicate training data (e.g., a visual style or music sample)
• Generated work mimics a celebrity voice, likeness, or persona

In 2023 and 2024, creators and rightsholders filed lawsuits against AI developers for training on books, images, and voice models without permission. These cases—some ongoing—may shape future obligations for data transparency and licensing in AI training.

IV. International Perspectives: Divergence in Global Policy

The global IP landscape is rapidly evolving:

• UK: Grants copyright to AI-generated works under a “computer-generated works” provision—but ownership defaults to the person who made “arrangements.”
• EU: Exploring AI Act provisions but maintains human authorship under copyright law.
• China: Moving toward protection of certain AI-generated content under unfair competition laws.

Implication: A lack of harmonized international rules could spark forum shopping and raise enforcement challenges across jurisdictions.

V. The Future: Legislative Reform or Legal Limbo?

Pressure is mounting on U.S. lawmakers to address the AI authorship gap. Key proposals under discussion include:

• A new sui generis right for AI-generated works (akin to database rights in the EU)
• Clarified disclosure obligations when using AI in commercial content
• Updated DMCA rules for algorithmic sampling and source attribution

Until reforms are codified, creators and companies must navigate a gray zone—where IP risk management matters as much as innovation.

Conclusion: Creators Without Consciousness, Laws Without Precedent

AI has upended the traditional model of authorship, challenging the legal system to rethink how we assign credit, ownership, and liability in creative industries. For now, the law remains clear on one point: machines can’t be authors. But as generative tools become more autonomous and pervasive, the distinction between human and machine creativity may prove harder to draw.

Until then, the art is artificial—but the legal risks are very real.