Sarah Silverman and Artists Sue OpenAI and Meta: Why

Sarah Silverman and other authors sued OpenAI and Meta because they say the companies used copyrighted books to train generative AI systems without permission or payment. Their lawsuits argue that this copying violated copyright law; the companies have disputed key claims, and the cases raise unresolved questions about how copyright applies to AI training.

What prompted the lawsuits

In July 2023, comedian and author Sarah Silverman joined writers Christopher Golden and Richard Kadrey in filing separate proposed class-action lawsuits against OpenAI and Meta in federal court in California. The OpenAI case named OpenAI entities; the Meta case named Meta Platforms. The authors said their books had been copied and used in building the companies’ language models without their consent.

Silverman is also known as a performer, but the claims at issue concerned her writing. The complaints identified books she had authored, including The Bedwetter. The authors alleged that OpenAI’s ChatGPT and Meta’s LLaMA could produce summaries or other outputs reflecting material from their books, and that this was evidence the works had been included in training data.

The suits sought to represent broader groups of copyright holders, rather than only the three named authors. A proposed class action does not automatically cover everyone who might have a similar claim: a judge must certify a class, and the parties may contest who would qualify.

How AI training relates to the claims

Large language models are trained on large collections of text. During training, a system processes examples and adjusts numerical parameters so it can predict likely next words. The resulting model is not ordinarily a searchable library that simply retrieves a complete book on request. However, the training process can involve making copies of source material, and models may sometimes reproduce memorized passages or closely resemble material in their training data.

The authors’ complaints alleged that copyrighted books were obtained from unauthorized online sources, including so-called shadow libraries, and copied for training. Those allegations are claims in litigation, not findings that every named work was used in a particular model or that every stage of the companies’ data handling was unlawful.

There are two related but distinct questions: whether copyrighted works were copied during data collection or training, and whether a particular model output infringes a copyright. The lawsuits focused on alleged copying and training, while the authors also pointed to outputs they said reflected their books. Establishing one issue does not automatically settle the other.

What the authors say was wrong

In broad terms, the authors argued that their books were used to develop commercial AI products without authorization, licensing fees, or credit. Copyright owners generally have exclusive rights over certain uses of their works, including reproduction, subject to legal exceptions and limitations. The lawsuits asked the court to determine whether the alleged copying fell outside those protections.

The complaints also raised claims connected to the alleged removal or absence of copyright-management information, and sought remedies including damages and court orders. The precise claims and requested relief depend on each case and can change as litigation proceeds. A filing is not proof of wrongdoing, and plaintiffs must establish the elements of their claims.

For writers, the dispute is about more than whether a model can imitate a style. A writing style by itself is generally different from copying protected text, while a distinctive passage or a substantial reproduction can present a different question. Authors have also raised concerns about the economic effect of AI systems that can generate text in response to prompts, but arguments about market harm require evidence and legal analysis.

What OpenAI and Meta have argued

AI developers have generally argued that training on publicly available material can be transformative and may qualify as fair use under U.S. copyright law. Fair use is a case-specific defense, not a blanket permission to use anything found online. Courts commonly weigh factors such as the purpose and character of the use, the nature of the copyrighted work, how much was used, and the effect on the work’s market.

The companies have also challenged aspects of the authors’ cases, including whether the complaints adequately connect particular works to particular models or outputs. Such arguments are part of the adversarial process; they should not be mistaken for a final ruling that training is lawful or unlawful. The central legal questions include what copies were made, how they were used, whether an exception applies, and what proof is required.

Why the cases matter beyond these authors

The lawsuits are part of a wider debate involving authors, visual artists, music publishers, news organizations, and AI developers. Their interests overlap but are not identical: text models, image generators, and music systems may use different datasets and produce different kinds of outputs. A decision in one case could offer guidance without automatically resolving every other dispute.

• For authors and publishers: the cases may affect whether training licenses, opt-out systems, or other compensation arrangements become more common.
• For AI companies: courts may clarify what evidence is needed to show that protected works were used and whether a particular use is fair.
• For users: the outcomes could influence which services and outputs are available, but the lawsuits do not mean that every AI-generated response is infringing.

It is also important not to confuse copyright in a book with ownership of a general idea, topic, or writing style. Copyright typically protects original expression, not facts or ideas as such. Whether a generated answer crosses the line depends on the material, the output, and the applicable law—not simply on whether an AI system was involved.

What has—and has not—been decided

The filings were made in 2023, and the litigation has involved procedural motions and evolving claims. Court proceedings can narrow a case, dismiss particular claims, or allow others to continue without resolving the underlying legal question for the entire technology industry. Readers should distinguish a judge’s ruling on whether a claim may proceed from a final decision after evidence and legal arguments are considered.

These suits do not, by themselves, establish that OpenAI or Meta unlawfully used every author’s work, nor do they establish that all AI training is fair use. The answer depends on facts and legal rulings. Because lawsuits can change through amendments, appeals, settlements, or later decisions, a current court docket or reliable recent reporting is the best source for the latest procedural status.

Frequently Asked Questions

Why did Sarah Silverman sue OpenAI?
She joined a 2023 lawsuit alleging that OpenAI copied her copyrighted book and other authors’ books without permission to train AI models, and seeking legal remedies.

Why did Sarah Silverman sue Meta?
She and the same co-plaintiffs filed a separate case alleging that Meta used copyrighted books, including works by the plaintiffs, in training its LLaMA models without authorization.

Did a court rule that the companies stole the books?
A lawsuit’s allegations are not a court finding. Whether particular copying infringed copyright depends on evidence, legal defenses, and the course and outcome of the litigation.

Is AI training automatically fair use?
No. Fair use is assessed case by case under multiple factors. The fact that a system learns patterns or serves a new purpose does not settle the legal question on its own.

Can an AI output infringe a book’s copyright?
It can raise copyright concerns if it reproduces protected expression, but not every similar answer or shared idea is infringement. The specific text and circumstances matter.

Related

• Elden Ring Multiplayer Not Working? Fixes That Work
• Epic Games Introduce Unreal Engine?
• Business IT Support for Remote Work: A Guide