In 2025, a federal court ruled that AI-generated images have no copyright protection because they lack human authorship. A few weeks later, a different federal court suggested that if a human made creative decisions directing the AI, copyright protection applied. The contradiction is not a legal accident. It’s evidence of a genuinely unresolved question: in an age of AI-generated content, who owns creative work?
The Three-Way Conflict
- Artists say: AI companies trained on our work without permission. We deserve compensation. Our creative autonomy is threatened by AI that imitates our style.
- AI companies say: We trained on publicly available data. The models don’t reproduce work—they synthesize patterns. Users make creative decisions about prompts. This is fair use.
- Consumers say: I want to use AI to create. The legal uncertainty makes it hard to build products and businesses on AI generation.
What Courts Are Actually Deciding
Multiple lawsuits in 2024-2025 established several precedents: 1) AI-generated work with no human creative input has no copyright. 2) If a human gave significant creative direction, copyright might apply. 3) Training models on copyrighted work is fair use if the output is sufficiently transformed. 4) AI companies may owe compensation to artists for training data use.
The problem: these rule together are contradictory and geographically inconsistent. What’s legal in the US may be illegal in Europe. What’s fair use in California might not be in New York.
The Business Model Crisis
The uncertainty makes it hard to build stable businesses on AI generation. A company offering AI-generated stock imagery faces copyright liability if they used copyrighted images for training. A music production company using AI generators faces questions about who owns the generated tracks. Every use case sits in a legal gray zone.
The Likely Outcome
Regulation will eventually create clarity. The most likely path is: 1) Licensing becomes required for training on copyrighted work. 2) Copyright extends to AI-generated work if the generator received significant creative direction from a human. 3) Orphan works (uncontacted creators) get handled through compulsory licensing and compensation funds. 4) Fair use narrows as courts establish that training on copyrighted work isn’t automatically fair use.
Until this happens, the creative industry faces a period of uncertainty that favors large companies (who can afford litigation) over creators and small startups.