Technology
Courts navigate decades-old copyright statutes in artificial intelligence training disputes
Recent rulings highlight judicial divergence over fair use protections as judges apply 1976 legal frameworks to generative AI.
The short version
- U.S. courts are delivering mixed rulings on whether using copyrighted works to train artificial intelligence models constitutes fair use under copyright law established in 1976.
- Judges have drawn distinctions based on whether training creates a transformative product or directly competes with original material, penalizing unauthorized source acquisition while permitting certain training mechanisms.
- The broader legal framework remains unresolved as numerous lawsuits proceed through the courts without updated statutory guidance.
Key facts
- Artificial intelligence models powering major chatbots rely on training datasets containing hundreds of millions of books, articles, and research papers, frequently without the knowledge or explicit consent of the authors.[TechCrunch]
- Judge William Alsup ordered Anthropic to pay a $1.5 billion settlement for downloading texts from unauthorized shadow libraries, while finding that the technical process of training AI models on the works to produce new content was lawful.[TechCrunch]
- In a separate case, Judge Stephanos Bibas ruled that Ross Intelligence did not qualify for fair use protections when training an AI legal platform using Thomson Reuters content because the system directly competed with the original work and lacked a transformative purpose.[TechCrunch]
- In Thaler v. Perlmutter, a court determined that works created entirely by artificial intelligence cannot receive copyright protection.[TechCrunch]
- Current United States copyright statutes date back to 1976, requiring judges to interpret 50-year-old standards for modern algorithmic training methods.[TechCrunch]
What remains uncertain
- Legal experts note that it remains uncertain how higher courts will ultimately reconcile diverging standards regarding market competition and transformative use in AI training cases.[TechCrunch]
- It remains legally and technically unresolved how courts will verify and quantify the percentage of AI assistance used in partially machine-generated creative works.[TechCrunch]