Ninth Circuit Limits DMCA Claims Against AI Code Generators

Court narrows DMCA liability for AI training

The Ninth Circuit Court of Appeals has issued a significant ruling in Doe v. GitHub that clarifies when AI coding tools can face liability under the Digital Millennium Copyright Act. The court rejected an interpretation that would have exposed AI developers to claims whenever their tools generated code resembling existing copyrighted work without including attribution.

The case centered on Section 1202(b) of the DMCA, which governs copyright management information (CMI)—details like author names, copyright notices, and licensing terms attached to copyrighted works. Plaintiffs argued that when AI coding tools trained on publicly available code produced similar output without the original CMI, that constituted a DMCA violation.

The court disagreed. In its opinion, the Ninth Circuit stated that to establish a claim under Section 1202, plaintiffs must demonstrate that defendants "removed or altered CMI from copies of existing protected works." Simply showing that a new work resembles an existing one but lacks CMI is insufficient without evidence of actual removal or alteration.

Why it matters

This ruling prevents a narrow anti-piracy provision from becoming a sweeping liability trap for AI development. Had the court ruled otherwise, every AI-generated code snippet that resembled existing work could have triggered DMCA claims, creating massive legal uncertainty for developers, digital services, and individual users. The decision preserves the distinction between creating new works and actively stripping copyright information from existing ones—a critical boundary for the AI industry.

Industry groups shaped the outcome

The Computer & Communications Industry Association (CCIA) and the Chamber of Progress filed a joint amicus brief arguing that imposing liability based solely on similarity would transform Section 1202(b) beyond its intended scope. Their brief emphasized that the provision prohibits removing or changing CMI from existing works, not failing to include it on new creations.

The court adopted this position, maintaining that the DMCA's CMI provision focuses on preventing infringement through alteration or removal of copyright information from existing copies, not on policing whether new works carry forward attribution.

Boundaries remain

The ruling does not provide blanket immunity for AI-generated code. The court acknowledged that substantially similar works can still face DMCA scrutiny, and similarity between works may serve as circumstantial evidence of CMI removal. Copyright and DMCA liability remain possible depending on specific circumstances.

What the decision does establish is that resemblance alone, without proof of actual removal or alteration of CMI from an existing work, cannot support a Section 1202(b) claim. This distinction preserves the provision's original purpose while preventing its expansion into a general-purpose copyright enforcement mechanism.

Details of the ruling were first reported by CCIA.