AI Copyright Lawsuits: Fair Use, Market Dilution, and Key Cases with Pam Samuelson

Understanding Copyright and Fair Use in AI

• 🔑 Copyright law grants authors exclusive rights, including the right to reproduce their work, but includes a fair use defense.
• 💡 When copyrighted works are used as training data for generative AI, it involves making reproductions, potentially leading to claims of infringement.
• ⚖️ The fair use analysis typically hinges on the transformativeness of the use and its effect on the market for the original work.

The Supreme Court's Warhol v. Goldsmith Decision

• 🖼️ The Warhol v. Goldsmith case involved Andy Warhol's use of Lynn Goldsmith's photograph of Prince, leading to a Supreme Court ruling that the commercial licensing of Warhol's artwork as a magazine illustration substituted for a market Goldsmith occupied.
• ⚠️ This case highlighted the significance of market substitution and commercial use in fair use determinations, though its broader impact on fair use doctrine is debated.
• 🏛️ The Supreme Court's decision emphasized that copyright's primary beneficiary is the public, aiming to promote progress in science and the arts.

Key AI Copyright Lawsuits: Bartz v. Anthropic

• 🤖 In Bartz v. Anthropic, authors sued Anthropic for allegedly using their copyrighted books to train AI models.
• 🚀 Judge Alup ruled that using copyrighted works as training data for constructing an AI model is transformative fair use, as it's for a non-expressive purpose (model construction) rather than consuming the expression.
• 🚫 However, the court found that using illegally downloaded books for training data was not fair use, leading to a proposed settlement.
• 🌍 This ruling aligns with some aspects of the EU's text and data mining exceptions, which permit non-profit and opt-out commercial use of lawfully acquired copyrighted material for training.

Kadrey v. Meta and Market Dilution

• 💻 In Kadrey v. Meta, similar claims were made against Meta, but the court focused differently, dismissing claims related to lost sales due to Meta's system guardrails against direct recitation.
• 📉 The court also rejected the argument that AI companies owe licensing fees for using works as training data, deeming it a market authors don't control.
• 🤔 A novel theory of market dilution was raised, suggesting AI outputs could satisfy demand for original works, but the court found insufficient evidence to support this claim.

Other Cases and Policy Considerations

• 📂 The DO v. GitHub case involves claims of removing copyright management information from open-source code used in AI training, focusing on whether identical code output constitutes a violation.
• ⚖️ The Thomson Reuters v. Ross Intelligence case, influenced by Warhol, found that an AI program offering similar services to Westlaw's tool constituted infringement due to market substitution.
• 📝 The U.S. Copyright Office issued a report stating that some AI uses might be fair use depending on factors like transformativeness and commerciality, and introduced the concept of market dilution.
• 🏛️ The dismissal and reinstatement of the Register of Copyrights, Shira Perlmutter, highlight ongoing legal and administrative complexities surrounding AI and copyright policy.