OpenAI appeals a court order in The New York Times case requiring it to keep ChatGPT outputs indefinitely, saying it conflicts with its user privacy commitments
OpenAI is appealing an order in a copyright case brought by the New York Times that requires it to preserve ChatGPT output data indefinitely …
Context & Ripple Effects
The appeal follows OpenAI's earlier effort to block the May 13 preservation order, arguing that retaining even deleted chats would undermine user privacy commitments. That dispute puts ChatGPT conversation data at the center of the evidence fight in the Times copyright case, rather than limiting the case to training-data claims.
The conflict is already framed by sharply contested allegations over how infringement evidence was obtained, including OpenAI's attempt to dismiss parts of the Times suit and the Times' subsequent rejection of OpenAI's characterization of its testing.
First-order effects
- OpenAI must continue contesting a preservation requirement that it says overrides its deletion and privacy practices, while the appeal determines whether the company can avoid indefinite retention of ChatGPT outputs.
- The New York Times gains a route to preserve potential evidence about ChatGPT's responses, strengthening its ability to test claims that the system reproduces its work if the order stands.
Second-order effects
- The case increases the tension between AI providers' data-minimization promises and litigation-discovery obligations; providers may face pressure to make retention, deletion, and legal-hold terms more explicit to users.
- Copyright plaintiffs gain an incentive to seek broad output records as evidence, while AI defendants will push for narrower, privacy-protective discovery protocols—the dispute OpenAI had already raised in its bid to block preservation of all user logs.
Third-order effects
- If courts increasingly treat model outputs and user interactions as discoverable evidence in content lawsuits, AI companies' operational data-governance choices could become a consequential part of copyright risk management.
- The enduring policy question is whether evidence preservation can be reconciled with meaningful deletion commitments; outcomes will help define the permission boundary for data generated on AI platforms.
The trend: This is one point in the broader shift from disputes over AI training data toward litigation over the outputs, records, and user-data controls needed to prove infringement claims.