OpenAI Mandated to Safeguard ChatGPT Conversations

OpenAI has formally challenged a court order mandating the retention of user conversations, which forms part of an ongoing copyright infringement lawsuit filed by The New York Times and other publishing entities. The publishers contend that ChatGPT’s responses often include their proprietary content to such an extent that users are accessing this information instead of their paid services.

The plaintiffs maintain that conversations deleted from ChatGPT could reveal instances where users accessed proprietary material through the service. During judicial discussions in January, Judge Ona T. Wang raised concerns that users aware of the case might delete conversations to obscure their activities. Although she rejected the publishers’ earlier request for a preservation order, she questioned OpenAI’s inability to anonymize and segregate data for users who preferred deletion, a point that OpenAI did not sufficiently address, prompting her order issued on May 13.

Judge Wang’s order stipulates that OpenAI must preserve and segregate all output log data that would typically be deleted moving forward, regardless of whether such removal occurs at a user’s request or due to various privacy laws requiring such actions.

By default, ChatGPT retains user interactions to enhance its AI model, though it offers settings to disable this feature, allowing users to have their conversations forgotten. Additionally, it includes a temporary chat mode that deletes exchanges upon conclusion.

In its objection to the order, OpenAI raised concerns about user privacy, asserting that compliance would undermine its commitment to user control over the retention and use of their conversation data. The company emphasized that ongoing enforcement of the preservation order compromises the privacy safeguards it has diligently instituted.

OpenAI argued that there is no substantial evidence suggesting that deleted conversations contain more of the publishers’ content and cautioned that users often discuss sensitive topics within these interactions, expecting that such details will not be retained.

Moreover, OpenAI indicated that modifying the data retention framework would require several months of engineering efforts.

The lawsuit’s origins involve three publishers—The New York Times, the New York Daily News, and the Center for Investigative Reporting—who initially filed separate lawsuits against OpenAI for copyright infringement, which were later consolidated. OpenAI has maintained that its use of content falls under fair use provisions, asserting that its AI model transforms original material into tokens, which are then combined with other information to generate user responses.

Despite conversations being deleted, ChatGPT retains a separate memory of user-provided details, which enhances its personalization capabilities. Users have the option to manage these memories, either by disabling references or deleting them entirely.

It is imperative for users to exercise caution when sharing information with online platforms, particularly AI services, recognizing the fluid nature of conversational data. Prudence is advised, especially concerning the sharing of sensitive details that might be regretted later.