Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
OpenAI says the indefinite, forward-looking obligation to preserve certain ChatGPT output logs ended on September 26, 2025. That date comes from OpenAI’s October 22, 2025 case update, not from a published terminating order identified here. The change ended the special preservation requirement; it did not mean every retained record was immediately erased, that all historical data was released, or that the New York Times copyright lawsuit was over.
What the order required
On May 13, 2025, U.S. Magistrate Judge Ona T. Wang directed OpenAI to preserve and segregate output-log data that otherwise would have been deleted. The direction covered data marked for deletion at a user’s request as well as data scheduled for deletion under OpenAI’s ordinary systems and legal obligations.
Judge Wang described the order as a litigation-preservation measure. The parties were litigating copyright claims involving the New York Times and other publishers, including questions about whether ChatGPT could reproduce publisher content. Preserving logs was intended to protect potentially relevant evidence during discovery. It was not a ruling that OpenAI infringed copyright, nor a general privacy or retention rule for every OpenAI product and account.
Recommended Free Tools
“Accordingly, OpenAI is NOW DIRECTED to preserve and segregate all output log data that would otherwise be deleted on a going forward basis until further order of the Court,” Judge Wang wrote in the May 13 order.
When did the special retention obligation end?
OpenAI’s October 22, 2025 update says that “our obligations under the earlier order ended on September 26, 2025.” The date should therefore be understood as OpenAI’s account of when the forward-looking, indefinite preservation requirement ended. The precise text of the court order ending that obligation is not established by the materials summarized here.
OpenAI said that, after the obligation ended, its ordinary deletion processes would resume. In that update, it described the following standard timing:
| Data category | OpenAI’s stated ordinary handling after September 26, 2025 | Important qualification |
|---|---|---|
| Deleted ChatGPT conversations | Automatic deletion within 30 days | OpenAI’s October 22, 2025 description of standard practice; legal and security exceptions may apply. |
| Temporary Chats | Automatic deletion within 30 days | Same dated OpenAI description and exceptions. |
| API data | Automatic deletion after 30 days | Applies to the standard practice OpenAI described at that time, not necessarily every account configuration or legal hold. |
A 30-day period is not the same as instantaneous erasure. It also does not override a separate legal hold, security requirement, backup process, or another applicable exception.
What data remained under a separate hold?
OpenAI said it continued to securely store a limited historical set of user data from April through September 2025. According to OpenAI, that material was locked down and accessible only to a small, audited legal and security team for legal obligations. OpenAI also said the data would not be turned over to the Times, the court, or anyone else “at this time.”
This historical set is legally and operationally different from the forward-looking order. The May order required preservation of records that would otherwise be deleted going forward. OpenAI’s later description concerns a limited period that had already been retained and remained subject to a separate legal hold.
Two periods, two statuses
| Period or set | Status | Who established or described it? |
|---|---|---|
| From May 13 until September 26, 2025 | Forward-looking preservation and segregation of output logs that otherwise would have been deleted, pending further court action. | Judge Wang’s May 13, 2025 court order. |
| April through September 2025 historical data | Limited data that OpenAI said remained securely stored under a separate legal hold. | OpenAI’s October 22, 2025 update. |
Did the order mean deleted chats were handed to the New York Times?
No such blanket conclusion follows from the order or OpenAI’s update. The preservation requirement concerned keeping potentially relevant records for discovery. OpenAI later said the limited historical data under legal hold would not be turned over to the Times, the court, or anyone else “at this time.” That statement is attributed to OpenAI and is temporally qualified; it is not a permanent ruling about every record or every future discovery decision.
The existence of preserved logs also does not establish that any particular user’s conversation was used, produced, or disclosed. The materials do not establish a population-wide count of affected users or that every conversation was handled identically.
How the court proceedings developed in 2025
May 16: reconsideration denied without prejudice
On May 16, Judge Wang denied OpenAI’s request to reconsider the preservation order without prejudice. The court identified unresolved questions, including whether logs marked for deletion differed from logs already retained and whether the breadth of preservation was proportional to the litigation.
Best Value
May 29: sampling ordered
On May 29, the court directed the parties to begin sampling 30-day tables of consumer output logs. It again denied reconsideration and deferred further preservation-related discovery until the sampling was complete. These were procedural steps about evidence and discovery, not a merits ruling on copyright infringement or fair use.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Did ending the data order end the lawsuit?
No. A discovery-preservation order and the underlying copyright case answer different questions. The preservation order governed how potentially relevant records were kept while the parties litigated. It did not decide whether training on publishers’ material infringed copyright, whether a fair-use defense applied, or whether any party was liable.
The broader litigation also continued after September 2025. A later memorandum order dated August 6, 2026, by Judge Sidney H. Stein addressed contributory-copyright theories and trademark-dilution claims, including dismissal with prejudice of certain claims and denial of proposed amendments against Microsoft. That later order is a separate development and is not the order that ended the 2025 forward-looking retention obligation.
What the “20 million logs” figure does—and does not—mean
OpenAI’s 2026 case-status material refers to a later production dispute involving 20 million retained, de-identified consumer ChatGPT output logs. That figure concerns a separate production issue. It should not be treated as the number of users affected by the May 2025 preservation order, the number of chats retained under the historical legal hold, or proof that all ChatGPT conversations were preserved.
What deleting a ChatGPT conversation means in practice
Based on OpenAI’s October 22, 2025 description, deleted ChatGPT conversations and Temporary Chats ordinarily returned to automatic deletion within 30 days after the special obligation ended, while API data was ordinarily deleted after 30 days. “Ordinarily” matters: legal holds and security or legal exceptions can change the timeline, and the update does not promise identical treatment for every data category, product, account, backup, or future policy.
Quick Recap
- The special court-ordered, forward-looking preservation requirement ended on September 26, 2025, according to OpenAI.
- OpenAI said a limited April–September 2025 historical set remained under a separate legal hold.
- Those statements do not show that every retained item was immediately erased or transferred.
- The copyright litigation and other discovery disputes remained separate and procedurally active.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.

