Summary
- OpenAI tries to reverse a court order demanding the handover of 20 million anonymized chat logs from ChatGPT.
- The court order states that privacy will be protected by exhaustive de-identification and other safeguards.
- The original lawsuit accuses OpenAI of improperly using news articles to train ChatGPT.
OpenAI is trying to reverse an order forcing it to hand over 20 million anonymized chat logs from ChatGPT. The company filed the request yesterday (12/11), arguing that the delivery of data violates users’ privacy.
In the request made to the court, OpenAI states that “99.99%” of the transcripts, requested by New York Times in a copyright case, are unrelated to the case. The company warned that the order affects anyone in the world who has used ChatGPT in the past three years, who now “must face the possibility of their personal conversations being turned over to the Times.”
Evidence of illegal reproduction of content
The original lawsuit accuses OpenAI of improperly using millions of news articles to train the models that power ChatGPT. It is worth remembering that the NOW does not completely oppose the use of content for AI training — as long as they are paid for it, as in the agreement reached with Amazon.
In this case, the vehicles argue that the 20 million chat logs are necessary to:
- Determine whether ChatGPT is, in fact, reproducing copyrighted content;
- Counter OpenAI’s claim that newspapers “hacked” the chatbot to fabricate evidence.
A spokesperson for NOW disagrees about users’ privacy being at risk, and claims that OpenAI’s blog post about the case “purposefully” misleads users and “omits the facts.”
According to him, the court order requires OpenAI itself to provide a sample of chats “anonymized by OpenAI itself”, protected by a legal order.
What did the court decide?
The original order was issued by Judge Ona Wang. The judge stated, in a decision, that users’ privacy would be protected by the “exhaustive de-identification” that OpenAI would carry out on the data, in addition to other safeguards.
The deadline set by the judge for OpenAI to deliver the transcripts ends this Friday (11/14). In Brazil, the company faces a similar process, brought by Folha de S.Paulo in August this year.
Source: https://tecnoblog.net/noticias/openai-tenta-reverter-ordem-para-entregar-20-milhoes-de-conversas-no-chatgpt/
