AI News

OpenAI’s 2025 ChatGPT Retention Order: Ended Obligations and Historical Hold

Conceptual illustration of a holographic ChatGPT shield, a gavel and a dissolving server
Conceptual illustration of retention and legal oversight; not a court photograph, a server deletion test or evidence that particular data was erased.

The Ended 2025 Retention Order and the Data Still Under Hold

OpenAI’s October 22, 2025 update said its obligations under the earlier order to retain consumer ChatGPT and API content indefinitely had ended on September 26. The company said it had returned to standard retention practices while keeping a limited set of historical data under legal hold. That distinction matters when interpreting what happens to a deleted chat.

This article originally covered October 2025 reporting about the order’s termination. It now incorporates OpenAI’s later October 22 statement. It preserves the original publication date and makes no claim about subsequent court rulings, today’s account settings or a test of OpenAI’s deletion systems.

The Background: When Privacy Met Copyright Law

Engadget’s October 11, 2025 report described the underlying copyright lawsuit as beginning in late 2023. The New York Times alleged that OpenAI had used its content without compensation. Those are allegations reported in the litigation, not findings of infringement established by this retention update. Related historical coverage: The New York Times filed a lawsuit.

Mashable’s October 12 report named The New York Times, The Intercept, Alternet and its parent Ziff Davis among publishers pursuing copyright claims involving OpenAI and Microsoft. Mashable disclosed its parent’s involvement. This historical account does not verify the later status of any plaintiff’s case.

The Controversial Preservation Order

Mashable reported that the May 13, 2025 order required OpenAI to “preserve and segregate all output log data that would otherwise be deleted on a going-forward basis.” OpenAI’s June statement described the affected consumer and API products. That was a specific preservation obligation, not proof that every conversation in every OpenAI product was covered. Related historical coverage: OpenAI was directed.

In its June 5 statement, OpenAI said the order affected ChatGPT Free, Plus, Pro and Team, along with API use without the applicable Zero Data Retention agreement. It said ChatGPT Enterprise, ChatGPT Edu and API customers using ZDR endpoints under its amendment were excluded. Those historical exclusions do not establish that Enterprise or Edu retains no data under ordinary policies.

OpenAI’s statement described deleted chats as removed from the account immediately and scheduled for deletion from its systems within 30 days, subject to legal or security exceptions and other stated conditions. Removing a conversation from a user’s view is therefore separate from verifying its removal from every retained system.

OpenAI Fights Back

OpenAI opposed the preservation requirement. Its June statement called the demand an overreach and said it conflicted with the company’s privacy commitments. That describes OpenAI’s position in the dispute.

In the June statement, COO Brad Lightcap said the demand conflicted with OpenAI’s privacy commitments and weakened privacy protections. The company said it had sought reconsideration and appealed to the District Court Judge. Its later October update is the relevant passage for the end of the earlier obligation. Related historical primary statement: blog post published in June 2025.

PCMag’s October 11 report said Sam Altman had described the data request as “unconscionable.” That quotation is attributed to PCMag’s historical account; this review did not independently recover the original X post.

User Outcry and Privacy Concerns

Conceptual illustration of concerned people holding phones beneath an imagined shadow
Conceptual illustration of privacy concerns; not a photograph of affected ChatGPT users or evidence of their reactions.

PCMag reported concern from some ChatGPT users about the order. The broader issue is easy to understand: people who delete a sensitive conversation need to know whether a legal obligation can still require its retention. Related reporting and Kingy.ai coverage: OpenAI; outcry from the ChatGPT user community.

A chat may contain a draft email, a health question or personal information. A deletion control should be understood alongside the service’s retention conditions. This report cannot promise that any particular conversation has been erased from every system.

The New Ruling: What Changed?

Engadget and Mashable reported an order filed on October 9, 2025 that ended the earlier going-forward preservation requirement. OpenAI’s October 22 update said the obligations under that earlier order ended on September 26. These dates refer to the reported filing and the end of the earlier obligations respectively; this review did not inspect the court docket. Related historical coverage: Judge Wang approved a joint order.

For deleted ChatGPT conversations and Temporary Chats, the update said OpenAI had returned to automatic deletion within 30 days. The accompanying policy explanation retained legal and security exceptions. API retention also depends on the relevant endpoints, configurations and legal obligations described in the statement. This is a dated account of company policy, not a guarantee for an individual user’s data. Related Kingy.ai coverage: ChatGPT.

For the limited historical April–September 2025 data, OpenAI said it would keep the material securely under legal hold. It said the data was accessible only to a small, audited OpenAI legal and security team, could be used only to meet legal obligations, and would not be turned over to the Times, the court or anyone else at that time.

Earlier October secondary reports described preserved logs as accessible to plaintiffs and referred to flagged accounts. OpenAI’s later October 22 statement explicitly limited access and said the historical data would not be turned over at that time. This article withdraws its earlier unqualified claim that Times lawyers could comb through the retained logs. It does not independently resolve discovery rights or later access orders.

What This Means for Users

The supported change is the end of the earlier indefinite-retention obligation, subject to the historical hold described by OpenAI. The update does not establish immediate deletion, permanent erasure in every circumstance or a complete resolution of the copyright litigation.

PCMag’s headline described permanent deletion as restored. The company’s stated legal and security exceptions provide necessary context for that wording. This review has not tested account deletion, examined backups or audited OpenAI’s compliance. Related historical coverage: As PCMag points out.

Deletion, training permissions and saved personalization are separate questions for users to check in current documentation. This historical review did not verify today’s memory interface or the effect of its individual controls, so it does not provide a current step-by-step settings guide.

The Bigger Picture: AI and Copyright

The October retention update did not decide the underlying copyright claims. This report covers the dated retention statement and the contemporaneous reporting; it makes no claim about the later outcome of the lawsuit.

OpenAI described the lawsuit as baseless in its June statement. Publishers’ allegations and OpenAI’s response remain opposing positions. This report does not decide fair use or reduce that legal question to a technical description of tokenization.

Earlier reporting included a June 2025 Reuters report. The judicial quotation previously attributed to it has been withdrawn because neither that report nor the underlying opinion was available for verification here.

Privacy vs. Discovery: A Delicate Balance

Preserving evidence and disclosing it are separate steps. Privacy concerns depend on both the scope of what is kept and who can access it.

Relevant evidence may matter to litigants. The existence of a preservation duty, the scope of a hold and permission to disclose the retained material are separate questions. Keeping data does not itself establish that plaintiffs can read it.

Indefinite retention of deleted data raises a privacy concern. The company’s October statement narrowed its description of the continuing hold to limited historical data. Readers should keep that scope in view rather than assume an unlimited obligation covering every user and product.

The Industry Impact

The dispute raises a broader question for AI services that hold user content: how clearly do their deletion policies explain legal exceptions? This case alone does not establish what another provider stores, which claims it faces or how it would respond to an order.

Copyright questions need their own legal evidence. The retention statement does not establish a universal rule about fair use or predict licensing costs and the pace of innovation.

What Happens Next?

The earlier obligation ended according to OpenAI’s dated update, while a limited historical hold remained. This report does not confirm later additions to flagged accounts, new disclosure rights or whether any retained data was subsequently shared.

The copyright allegations remain outside the scope of this retention account. The October update does not establish a later judgment or damages award against OpenAI or Microsoft.

Lessons for AI Users

For users, the useful distinction is between the deletion control they operate, the provider’s stated retention policy and an exception that may apply to particular data.

Avoid treating a deletion promise as an assurance that sensitive information can never be retained or disclosed. OpenAI’s own statement includes legal exceptions. This report does not establish whether a specific conversation was used for training, exposed or made available in litigation.

Check current provider documentation for deletion, training permissions, Temporary Chats and saved memory. This historical article does not provide a tested guide to today’s account interface.

The Road Ahead

Conceptual illustration of an illuminated road with an OpenAI label, scales, gavels and a padlock
Conceptual illustration of privacy and legal questions; not a diagram of OpenAI’s storage systems or a prediction of the litigation’s outcome.

The practical lesson is to read deletion conditions together with any legal exceptions. The continuing historical hold described here should be kept distinct from the standard practices for new data. Related Kingy.ai coverage: AI.

OpenAI said its earlier indefinite-retention obligations ended on September 26, 2025 and standard practices resumed. It also described limited historical data that remained under hold. Both statements belong in an account of the change.

No Kingy.ai account-deletion test, legal investigation or later docket review was performed for this article. The historical hold and product exclusions are attributed to OpenAI, with secondary reports identified as such.

Sources