In a series of landmark rulings, multiple federal courts have dismissed claims under Section 1202(b) of the Digital Millennium Copyright Act against AI models, effectively allowing developers to use copyrighted material for training without attribution. These dismissals, seen in cases such as Raw Story Media v. OpenAI, Inc. and Andersen v. Stability AI, Ltd. present a significant legal hurdle for creators seeking recognition or payment for their contributions to AI model development, according to Eff.
Courts are largely dismissing copyright infringement claims against AI models, but the industry still creates a multi-billion dollar market for licensing training data, and legislators push for mandatory disclosure. A growing divide exists between current legal interpretations and evolving industry practices.
Based on current legal trends and emerging legislative efforts, companies will likely face increasing pressure to adopt transparent disclosure practices for AI-generated content, even as direct copyright infringement claims remain challenging for individual creators to win.
The Current Legal Battleground for AI Content
A court dismissed claims in Kadrey v. Meta Platforms, Inc. stating Meta’s LLaMA models are not infringing derivative works, according to Eff. The ruling solidifies the judicial stance: AI models trained on existing content do not automatically create infringing derivative works.
However, the legal landscape includes specific vulnerabilities for AI developers. In Andersen v. Stability AI Ltd. a court allowed copyright claims to proceed if a model was trained on a plaintiff's work and generated similar artistic outputs when their name was used as a prompt, according to Eff. The specific carve-out provides a critical roadmap for creators to successfully challenge AI models, potentially opening a floodgate of targeted litigation against developers.
While broad claims of derivative infringement and DMCA violations against AI models are largely dismissed, specific instances where AI output directly mimics a creator's distinct style when prompted by their name may still find legal standing. The nuance suggests a future where successful litigation against AI developers will depend on demonstrating direct, attributable stylistic appropriation, rather than general use of training data.
