An AI image can look like a familiar artist’s work. That raises two different questions: Can the new image be copyrighted, and was the artist’s work used lawfully to make the model? section 4.11
Copyright is an exclusive right over a work of creative expression, such as an image, text, movie, or song. Its owner is typically the person authorized to copy, distribute, display, or perform the work. The right lasts for a limited time; afterward, the work enters the public domain. Copyright is one form of intellectual property, distinct from patents, trademarks, and trade secrets. section 4.11
Generative AI puts several boundaries of copyright under pressure:
- AI-generated output: The US Copyright Office says AI-generated media alone is not eligible for copyright protection. A work containing AI-generated elements might qualify if it includes sufficient human creativity; that judgment is case by case. section 4.11
- Training data: Artists’ and writers’ existing works may be included in model training, prompting unresolved questions about infringement. Fair use is part of this debate, but an exclusive right does not answer every question about permitted use. section 4.11
- Licensed reuse: Open source does not mean unrestricted use. Licenses can permit proprietary reuse or impose conditions such as attribution or noncommercial use. section 4.12
Code-generation dispute
Developers sued Microsoft, GitHub, and OpenAI over Copilot, a code-writing model trained on GitHub repositories. The parties dispute whether that use violates developers’ rights. Most copyright-related claims were dismissed, while claims concerning open source licenses and breach of contract remained to be decided. section 4.12
Copyright can protect creators while leaving room for reuse and new expression. Generative AI makes that balance harder to assess: training practices, human contributions to outputs, and licensing arrangements all matter, and important legal questions remain unresolved. section 4.11 section 4.15
