Key ex-OpenAI researcher subpoenaed in AI copyright case

Alec Radford, a researcher who helped develop many of OpenAI’s key AI technologies, has been subpoenaed in a copyright case against the AI startup, according to a court filing Tuesday.

The filing, submitted by an attorney for the plaintiffs to the U.S. District Court in the Northern District of California, indicated that Radford was served a subpoena on February 25.

Radford, who left OpenAI late last year to pursue independent research, was the lead author of OpenAI’s seminal research paper on generative pre-trained transformers (GPTs). GPTs underpin OpenAI’s most popular products, including the company’s AI-powered chatbot platform, ChatGPT.

Radford joined OpenAI in 2016, a year after the firm’s founding. He worked on several models in the company’s GPT series, as well as a speech recognition model, Whisper, and DALL-E, the company’s image-generating model.

The copyright case, “re OpenAI ChatGPT Litigation,” was brought by book authors including Paul Tremblay, Sarah Silverman, and Michael Chabon, who alleged that OpenAI infringed their copyrights by using their work to train its AI models. The plaintiffs also argued that ChatGPT infringed their works by liberally quoting those works sans attribution.

Last year, the Court dismissed two of the plaintiffs’ claims against OpenAI, but allowed the claim for direct infringement to move forward. OpenAI maintains its use of copyrighted data for training is protected under fair use.

Redford isn’t the only high-profile figure who attorneys for the authors are attempting to wrangle. Plaintiffs’ lawyers have also moved to compel the deposition of Dario Amodei and Benjamin Mann, both ex-OpenAI employees who left the company to start Anthropic. Amodei and Mann have fought the motions, claiming they’re overly burdensome.

A U.S. magistrate judge ruled this week that Amodei must sit for hours of questioning about the work he did for OpenAI in two copyright cases, including a case filed by the Authors Guild.