There is a great deal of litigation pending, starting with respect to AI-generated images and texts and whether such images and texts infringe upon copyrights held by original creators and holders of those copyrights. The most-watched case is pending in the federal court in the Northern District of California, and it is called Andersen v. Stability AI, Ltd. (N.Dist. Cal.). In that case, a number of plaintiffs have sued a number of companies that commercially provide AI image-generating software/web platforms for users. For now, the case does not involve AI-generated text. The plaintiffs claim that, in various ways, the AI programs violate U.S. copyright laws. The claims can be summarized as follows:

• Direct infringement — this is based on allegations that the AI companies downloaded or otherwise acquired copies of five billion copyrighted images without permission and used those images to train (“training images”) the AI program at issue and caused those images to be stored at and incorporated in the AI program as compressed copies; unauthorized acquisition and storage of copyrighted images can, under certain circumstances, be deemed copyright infringement.
• Inducing vicarious copyright infringement — this legal claim asserts that, by distributing the AI program via third-party downloads and use/deployment of the AI program, the defendants are inducing consumers and other third parties to engage in copyright infringement; inducement and/or facilitation by one party of infringement by another is a form of infringement under U.S. copyright laws.
• Unauthorized creation of derivative works — under U.S. copyright law, copyright holders can prohibit the creation of derivative works; the plaintiffs claim that the “output images” provided by the AI program to consumers are derivative works that constitute copyright infringement.
• Digital Millennium Copyright Act (“DMCA”) claims — the DMCA prohibits persons from removing or altering copyright management information and other aspects of digital images, such as removing a copyright notice, the name of, and other identifying information about, the author of a work, etc.; the plaintiffs claim that the original training images had such information removed before providing the images to the AI program.

After the Andersen case was originally filed, in the usual course of litigation, the defendants filed various written Motions to have the various legal claims dismissed. The only claim that the court allowed was the direct infringement claim. The court found that unauthorized acquisition and storage claims were adequately stated in the original complaint. See Andersen v. Stability AI, Ltd. (N.Dist. Cal. Oct. 30, 2023). See the link above. Note that for a Motion To Dismiss, the court must assume “as true” any facts that are pleaded in the Complaint. Whether such facts are actually true will be determined in the later stages of the case. The court granted the plaintiffs an opportunity to amend and rewrite their claims.

The plaintiffs proceeded to amend their Complaint. The Complaint retained the direct infringement, inducement, and the DMCA-related claims. The plaintiffs dropped any infringement claim that the output images were derivatives of the original works. New Motions to Dismiss were filed.

On August 12, 2024, the court issued its opinion on the new Motions To Dismiss. Since the direct infringement claim had already been approved, the court addressed the inducement and DMCA claims. The court agreed that the inducement claim was sufficiently pleaded based on the fact — taken as true — that the AI program that was downloaded and used contained compressed copies of training images. That fact alone was sufficient for the inducement claim to go forward. The DMCA claims were dismissed since the plaintiffs had no facts to support the claim that copyright management information had been removed by the defendants. The court also dismissed a number of state law-based claims.

Contact The AI and Internet Attorneys at Revision Legal

The Broader Landscape of AI Copyright Litigation

While Andersen v. Stability AI is among the most closely watched cases, it is by no means the only significant AI copyright lawsuit working through the federal courts. Several parallel cases raise overlapping and distinct legal theories that will collectively shape how copyright law applies to generative AI systems.

In Getty Images (US), Inc. v. Stability AI, Ltd. (D. Del.), Getty Images filed suit alleging that Stability AI scraped more than 12 million photographs from Getty’s database without authorization to train its Stable Diffusion model. Getty’s complaint focuses on the commercial scale of the scraping, the removal of copyright management information, and the fact that Stable Diffusion outputs sometimes include distorted versions of Getty’s watermark — a visible sign, according to Getty, that its images were used in training. That case directly implicates a large commercial library’s economic interests and may be more compelling on the direct financial harm element.

In the music space, a coalition of major record labels — including Universal Music Group, Sony Music Entertainment, and Warner Records — filed suit against AI music generation platforms Suno and Udio in 2024. Those cases center on whether large-scale training on copyrighted sound recordings infringes the rights of copyright holders, even when the AI outputs do not reproduce the originals verbatim. These music cases add nuance because the question of what constitutes copying in the context of latent-space diffusion models is distinct from the image and text contexts.

The Copyright Office’s Position on AI-Generated Works

The U.S. Copyright Office has issued important guidance clarifying the copyright status of AI-generated content. In February 2023, the Copyright Office reversed the registration of a comic book called Zarya of the Dawn to the extent the registration covered AI-generated images produced by Midjourney, while allowing the copyright in the human-authored text to stand. The Office reiterated its longstanding position that copyright protection requires human authorship — a requirement grounded in the constitutional text, which grants Congress the power to promote the progress of science and useful arts by securing rights to “Authors.”

In March 2023, the Copyright Office published registration guidance stating that it will not register works produced entirely by a machine without creative input from a human author. However, works that contain both AI-generated and human-authored elements may be registrable with respect to the human-authored portions. The Office requires applicants to disclose the use of AI-generated content and to identify the portions of the work that are human-authored.

Practically, this means that companies relying on AI-generated marketing copy, product images, or other creative outputs may not be able to obtain copyright protection for those works. This creates an asymmetric risk: the AI training data likely infringes existing copyrights (according to plaintiffs in pending litigation), while the AI outputs themselves may not qualify for new copyright protection. Businesses should plan accordingly and consider investing in human-authored creative work that can be registered and protected.

What Businesses Should Do Now

Given the fast-moving and uncertain state of AI copyright law, businesses using AI tools to generate or process creative content should take several protective steps. First, conduct an audit of all AI tools in use. For each tool, understand how the developer licensed — or did not license — training data. Some AI developers have published information about their training data sources and licensing practices; others have not. Using an AI system built on unlicensed training data exposes your business to potential secondary liability claims.

Second, review your contracts with AI vendors. Many vendor agreements contain indemnification provisions — or conspicuously lack them — with respect to intellectual property claims. If the AI system you are using generates infringing outputs, does your contract require the vendor to defend and indemnify you? If not, negotiate for that protection before you renew or extend the contract.

Third, monitor the litigation. The Andersen case and the music industry lawsuits will likely result in significant rulings over the next year or two. Businesses that stay informed and adjust their practices accordingly will be better positioned to comply with emerging legal standards.

For more information, contact the experienced AI and Internet Lawyers at Revision Legal. You can contact us through the form on this page or call (855) 473-8474.