OpenAI researcher feared 'optics' of piracy, filings say

Unsealed court filings in the authors' copyright case against OpenAI and Microsoft quote researcher Sam McCandlish. In 2020, he said he worried about the "optics" of the phrase "openai uses copyrighted data from sketchy russian website" appearing on Hacker News, the Authors Guild said.

The filings, unsealed as part of the multidistrict litigation over OpenAI's use of the books database LibGen, also quote then-research director Dario Amodei calling the site "a bit sketchier" as a training source, according to the Authors Guild. OpenAI deleted its LibGen files in 2022 under an internal effort called Project Clear, the filings say.

Policy director Jack Clark wrote in May 2020 that the company's work would create systems that "substitute for the labor of" people. He predicted OpenAI would "likely ignore" author complaints "and release anyway," according to the unsealed material. A later hire brought on to improve GPT's writing dismissed author concerns as "acceptable economic disruption," the Authors Guild said.

The Hacker News discussion of the filings drew 274 points and 245 comments within hours. Publishers Weekly separately reported on the same unsealed documents. Plaintiffs including John Grisham, George R.R. Martin and Jodi Picoult are seeking damages in the consolidated suit, filed in Manhattan federal court.