Art World Uprising: Understanding the Lawsuit Between Sarah Silverman, Other Talents Versus AI Giants

Artificial intelligence models like ChatGPT and Bard are trained using data created by humans. The more data they ingest, the smarter they become at imitating human intelligence and creativity. Big players in the AI industry, like OpenAI and Meta, have deployed large language models trained by scraping text and books available online to extract data.

Because of how LLMs are trained, it was inevitable there would be a conflict between copyright law and artificial intelligence. Now the chickens are coming home to roost, as Sarah Silverman and other artists are suing OpenAI and Meta over copyright infringement.

Why Sarah Silverman and Other Artists Are Suing Open AI and Meta

Ina class action lawsuit [PDF] filed in California, comedian Sarah Silverman and other writers (Christopher Golden and Richard Kadrey) seek to recover damages against OpenAI and Meta over copyright infringement. The lawsuit alleges OpenAI and Meta scraped copyrighted books from pirate websites to train their AI models. This is the equivalent of an AI model downloading its training datasets from Piratebay without compensating the authors.

Coincidentally, aseparate class action lawsuit [PDF] against OpenAI alleges the company used unauthorized private information to train ChatGPT. Google is also facing a similar lawsuit over allegedly using stolen data to train Google Bard. This is why you shouldmake it a habit to protect your personal information , though publishing work and private personal data are not the same.

What Are the Chances Sarah Silverman Could Win the Lawsuit?

Image Credit:freepik

Silverman and other artists claim that ChatGPT can accurately summarize their books when prompted. The complaint argues that wouldn’t be possible if the AI model didn’t have access to the copyrighted material. However, if ChatGPT was trained using billions of internet texts, it likely came across articles, comments, and social media posts discussing the books.

Furthermore, Meta disclosed where it acquired the books it used to train its AI model—the source was traced to an e-book torrent website. Similarly, the class action lawsuit against OpenAI also mentions illegal websites where OpenAI is suspected to have obtained the copyrighted materials, but OpenAI is yet to confirm its sources.

If it’s proven that OpenAI and Meta used illegal torrent websites to acquire copyrighted materials to train its AI models, Silverman may have a shot of winning the lawsuit. However, AI models are uncharted territory with no precedent for courts to rely on to make a ruling based on AI copyright infringement. In fact, that’s one of the reasons whythe EU proposed an AI Act.

What Is the Future of AI Models and Copyright Law?

We’re still in the early days of AI to know how it will adapt to copyright law. It’s even more complex trying to figure outwho owns the copyright to AI creation . But for human creators, regulations exist to protect them against someone else accessing their copyrighted material without compensation, consent, or credit. If the rules exist for humans, will they apply to AI models?

The EU parliament drafted the closest glimpse into the future of how AI models will comply with copyright law. If the EU AI Act is assented into law, AI models like ChatGPT and Bard will be required to publish all their dataset sources and copyrighted data used for training. This will help clear up any confusion if the AI models accessed copyrighted books, movies, music, and photos for training through illegal piracy websites.

The AI Class Action Lawsuit Ruling Will Set a Legal Precedent

Large language models can scrape all the corners of the internet for data used in training. But would they be liable for copyright infringement if they access illegal torrent websites to acquire the data? And if they do, can you prove it?

Regardless of the outcome, the class action lawsuits against the tech companies that own the most popular AI models will set a precedent that will be relevant in the future.

Also read: