OpenAI is fighting to keep Ilya Sutskever’s files secret in the process

According to the letter, lawyers for both sides met several times and agreed on a list of 24 custodians, but reached an impasse over those six names, which include the co-founder and former chief scientist Ilya Sutskever, Qiming Yuan Preform Data Coordinator, Tech. staff member Jong Wook Kim, technical staff member Shantanu Jain, former researcher Cullen O’Keefe and former science communicator Andrew Mayne. Mayne has written publicly about the importance of books as instructional data for the LLM.

Sutskever was instrumental in the orchestration firing CEO Sam Altman in November 2023 and has since left the company for start a competing AI model developer.

The lawyers say in the letter that they believe he has documents relevant to the trial. OpenAI, for its part, has not publicly said why it opposes including Sutskever and the other five in the discovery process. Now, lawyers for the Authors Guild are asking the judge to step in and compel OpenAI to include those names as custodians.

All six exhibits attached to the letter have been fully redacted at OpenAI’s request. The company said in court documents that it requested the redaction because the exhibits include proprietary source code as well as “discussions between OpenAI employees describing detailed processes for training and testing ChatGPT models.”

The process is one of many claiming that OpenAI infringed copyright when it used millions of books to train its AI models. Plaintiffs include authors such as George RR Martin, Jonathan Franzen, and David Baldacci.