Understanding Copyright Risks in AI-Generated Images

Understanding Copyright Risks in AI-Generated Images

Generative AI has revolutionized digital image creation, enabling users to generate images quickly from text prompts. However, according to a Bloomberg Law article by attorneys from the Hamilton Brook Smith Reynolds firm, using these AI-generated images poses copyright risks, as demonstrated by the recent Andersen v. Stability AI case in California. 

In that case, the court ruled that AI companies and users can be liable for copyright infringement if AI-created images closely resemble those in the program’s training set, including scraped public images.

AI models learn from massive datasets, and the result may replicate elements of copyrighted works, such as a Warhol-style image of Marilyn Monroe. To avoid copyright infringement, the authors suggest users should conduct thorough copyright searches to determine if their AI-generated image is similar to existing, copyrighted images. While AI-generated images aren’t entitled to copyright protection, users must be cautious.

The article also notes various search methods, including reviewing the AI’s training set, reverse image searches, and consulting attorneys, can help identify potential infringements. Though a copyright search can be difficult, building an “innocent infringer” defense is necessary, which can reduce damages in court if an infringement claim arises.

The innocent infringer defense allows courts to reduce statutory damages if the user can prove they unknowingly infringed. However, even with searches, some similar copyrighted images might go undiscovered, leading to possible litigation.

Companies and individuals using AI-generated images should proactively avoid legal risks, such as conducting thorough searches, seeking licenses, or consulting legal experts. This due diligence can minimize the chance of copyright infringement and associated damages.