ChatGPT will no longer retain deleted user conversations

Your deleted chats weren’t really gone

Imagine clearing your browser history, then learning a court ordered the company to preserve certain ChatGPT output logs that would otherwise have been deleted.

In May 2025, a magistrate judge directed OpenAI to preserve and segregate ‘output log data’ (the model’s generated responses) going forward, which meant some chats users deleted or set as temporary were held for legal discovery.

The policy shift impacted millions of people using free and paid versions of ChatGPT. Only certain enterprise and education customers were exempt from this controversial data retention rule, leaving most users unexpectedly exposed.

A courtroom battle over AI

The New York Times sued OpenAI, claiming it used copyrighted articles to train its AI models without permission. This legal action sparked a high-stakes battle about intellectual property in the age of artificial intelligence.

To gather evidence, the court ordered OpenAI to preserve all chat logs as potential proof. This legal demand immediately overrode the normal privacy promises the company had made to its vast user base.

Your privacy took a backseat

OpenAI fiercely argued that saving deleted chats violated user privacy, but the judge ruled against them. The court decided the lawsuit’s need for evidence was more critical than individual data concerns.

Because of the preservation order, OpenAI paused its routine deletion of certain chats and preserved output logs that otherwise would have been removed. Those preserved records remained available to plaintiffs as part of discovery until the court later narrowed the mandate.

What log preservation actually meant

The May 13 order used the phrase ‘output log data.’ i.e., ChatGPT’s responses and associated output records, as the primary category to be preserved.

Practically, these output logs can include metadata that links responses to sessions. The order targeted outputs rather than broadly redefining OpenAI’s entire retention regime for user inputs.

The news organizations meticulously searched these saved logs for clear examples of their copyrighted material appearing verbatim. They also looked for instances where the AI might have attributed false information to their reputable publications.

The order finally gets lifted

A joint stipulation filed with the court led to an order entered Oct. 9, 2025, terminating OpenAI’s ongoing obligation to preserve all output log data, and the parties’ filing says that the obligation ended effective Sept. 26, 2025, for most users, though previously preserved logs remain within the plaintiffs’ discovery set.

This change means the normal, automated deletion process for your old chats has now officially resumed.

But there’s a major catch

Conversations preserved before the cutoff remain part of the plaintiffs’ discovery. Those records are stored under legal hold and a protective order.

Reporting also notes some jurisdictional or contractual protections may limit access to data corresponding to users in certain territories (e.g., some EEA/UK/Switzerland scenarios), but preserved data generally remains available to plaintiffs for review under the court’s discovery process.

Furthermore, OpenAI must continue monitoring specific user accounts flagged by the news organizations during their review. This means a select group of people will still have their deleted chats saved and examined.

Who is still being watched?

If you operate a website or online domain that the plaintiffs have specifically flagged, your account remains under scrutiny. OpenAI is legally required to continue saving the deleted and temporary chats for these identified users. The list of monitored domains could potentially expand later.

This situation creates a two-tier privacy system where most users regain their data control, but some do not. Your professional or online presence could ultimately determine if your chats are now truly deleted after you remove them.

Why this lawsuit affects you

The outcome could completely reshape how AI companies operate and use online information for training. If the news organizations win, it might force AI firms to pay licensing fees for all their training data.

The landmark case also rigorously tests the legal boundaries of “fair use” in the new age of artificial intelligence. Your everyday AI interactions are directly at the heart of this multibillion-dollar legal confrontation.

More players joined the fight

Several other prominent publishers have joined the lawsuit against OpenAI and its partner, Microsoft.

Several news organizations have sued OpenAI in separate but related actions, including The New York Times, The Intercept, Raw Story, AlterNet, and a group of daily newspapers, among others. These actions were consolidated for some discovery purposes in the SDNY litigation.

The entire media and technology industries are watching this case’s progression very closely. Its outcome will likely set a powerful legal precedent for many future lawsuits involving AI and copyrighted content.

Microsoft tries to step aside

Interestingly, Microsoft is actively attempting to remove its AI tool, Copilot, from this specific litigation. The tech giant argues its product should be judged separately from OpenAI’s ChatGPT in this legal context. This strategic move highlights the complex and sprawling nature of the lawsuit.

It also shows that each defendant company is carefully strategizing to minimize its own unique legal and financial risks. The courtroom battle features numerous shifting legal alliances and individual defense strategies.

The stakes are incredibly high

News organizations allege AI tools directly threaten their business by replacing them as primary information sources. They also seriously fear reputational damage if chatbots repeatedly attribute false claims to their trusted brands.

Meanwhile, OpenAI faces potentially catastrophic financial damages that could reach billions of dollars. The immense pressure to settle the case out of court is growing, even from their own cautious insurance providers.

OpenAI’s normal privacy policy

Typically, when you delete a ChatGPT chat, it instantly vanishes from your account’s visible history. The company’s systems then schedule it for permanent, irreversible deletion from their servers within 30 days.

The “Temporary Chat” feature is specifically designed for automatic deletion within 30 days without any user action required. This option was created explicitly for people wanting more discreet and confidential interactions.

Sam Altman’s strong reaction

OpenAI’s CEO publicly criticized The New York Times’s extensive data request on social media. He powerfully called the demand for vast user logs “unconscionable,” highlighting the company’s deep frustration with the initial court order.

Legal representatives for OpenAI similarly accused the newspaper’s legal team of a massive “overreach.” They firmly believed the request fundamentally broke the trust carefully built with their global user community.

You still can’t recover old chats

Even with the special court-ordered retention, you could not retrieve your own deleted conversations. The preserved logs were kept in a separate, highly secure system accessible only for authorized legal review.

Deleting a chat still permanently removes it from your personal view inside the application. The unique court order did not change the fundamental user experience or functionality within the standard ChatGPT interface.

What about your uploaded files?

Files you upload into a conversation, like PDFs or images, are tied directly to that chat’s lifecycle. Whenever you delete the conversation containing your uploaded file, the file is also scheduled for automatic deletion. The same maximum 30-day timeline generally applies to both the chat and its files.

However, files uploaded into custom GPTs or shared projects are retained until you delete that entire GPT or project. Effectively managing your data within the platform often requires more than simply deleting a single chat history.

The big picture on AI privacy

This situation serves as a stark reminder that our conversations with AI may not be entirely confidential. CEO Sam Altman himself has openly warned that chats lack legal protection like attorney-client conversations. They could potentially be subpoenaed as evidence in various future lawsuits.

While this specific data order has ended, it establishes a powerful legal precedent for what can be compelled. It reminds us that our digital privacy can be surprisingly fragile and subject to complex legal demands.

Curious about how user feedback shapes AI? See why ChatGPT recently reversed its voice update.

How to protect your privacy now

For the highest level of privacy, consistently use the “Temporary Chat” feature for any sensitive topics. These chats are engineered to leave no trace in your history and are automatically deleted by the system. It remains the simplest built-in tool for conducting truly confidential inquiries.

Enterprise and education customers were exempt from the logging order due to their stronger, pre-negotiated data agreements. For regular users, simply being very mindful about what you share with any AI chatbot is your best personal defense.

Speaking of AI tools popping up in new places, find out how Claude is now joining the conversation in Microsoft apps.

Got questions about AI privacy? Share your thoughts in the comments, and if you found this helpful, pass it along.

Read More From This Brand:

• Apple takes on ChatGPT with Siri upgrade
• Former OpenAI researcher reveals ChatGPT’s errors
• ChatGPT launches parental tools to safeguard teens

Don’t forget to follow us for more exclusive content right here on MSN.

If you liked this story, you’ll LOVE our FREE emails. Join today and be the first to get stories like this one.

This slideshow was made with AI assistance and human editing.