OpenAI is pushing back against a federal court order that would force the company to save every single ChatGPT conversation — including chats users have explicitly deleted. The order stems from a copyright lawsuit brought by authors including Paul Tremblay and Sarah Silverman. They want those logs preserved as potential evidence. OpenAI calls the request "sweeping" and technically problematic.
Here's the uncomfortable truth. Most people click "delete" and assume the data is gone. I've done it myself. But in legal battles, "deleted" often just means "marked for eventual overwriting" — not actually destroyed. The court knows this. That's why they want OpenAI to freeze everything. And OpenAI is basically saying: that's not how our systems work, and doing this would break things.
This isn't just legal drama. It's a collision between user privacy expectations and the messy reality of AI litigation. Let's walk through what's actually happening.
Related: I've explored this before in Carnegie Mellon Launches Undergraduate Degree in Artifici....
Why Is a Court Demanding All ChatGPT Logs?
The short version: authors are suing OpenAI for copyright infringement. They claim ChatGPT was trained on their books without permission. To prove that, they need evidence showing the model can reproduce their copyrighted material. That evidence? User conversations where ChatGPT might have spit out passages from their work.
Here's where it gets tricky. The plaintiffs' legal team asked OpenAI to preserve all conversation data — every prompt, every response, every deleted chat — from the start of the lawsuit forward. The logic is straightforward: if users are generating copyrighted content through ChatGPT, those logs are the smoking gun. Delete them, and the evidence disappears.
Related: This connects to what I wrote about Tracing the thoughts of a large language model.
OpenAI's response, filed in late March 2025, argues this order is technically unworkable. According to their court filing, preserving deleted chats in a "frozen" state would require fundamentally redesigning their data deletion architecture. They'd have to build entirely new systems just to comply. And the cost? Significant. OpenAI hasn't put a public dollar figure on it, but they've called the burden "extraordinary."
I've worked with data retention systems before — not at OpenAI's scale, obviously — but even at smaller companies, "just save everything forever" is never simple. Data gets fragmented across servers. Deletion processes are baked into the infrastructure. You can't just flip a switch.
Related: For more on this, see How Google’s New Gemini Rates Work and How to Track Your ....
What OpenAI's Legal Argument Actually Says
OpenAI's opposition filing makes three main points.
First, they argue the order is overbroad. It covers all ChatGPT users, not just the ones who might be generating copyrighted material. That's hundreds of millions of conversations. Most have nothing to do with the lawsuit. OpenAI says this is like demanding a library preserve every book checkout record because someone might have photocopied a page.
Second, they claim deleted chats are genuinely deleted — not just hidden from users. According to their filing, once a user deletes a conversation, it enters a deletion pipeline and is permanently removed within 30 days. Preserving those chats would mean halting that pipeline entirely. For all users. Globally.
Third, they raise privacy concerns. Users delete chats for legitimate reasons. Maybe they shared sensitive personal information. Maybe they discussed a business strategy. Maybe they just changed their mind. OpenAI argues that preserving deleted conversations against users' wishes undermines the trust that ChatGPT is built on.
This last point is interesting because it's both a legal argument and a PR move. OpenAI is essentially saying: we're the ones protecting your privacy here. The court order threatens it.
The Technical Reality of "Deleted" Data
Let's talk about what "deletion" actually means in modern cloud systems. It's rarely instant. When you delete a ChatGPT conversation, the data doesn't vanish like a file going to the trash bin on your desktop. It gets flagged for removal. Background processes eventually overwrite the storage blocks. This can take days or weeks.
During that window, the data still exists. It's just not accessible through the normal interface. Forensic tools can sometimes recover it. This is why the plaintiffs want preservation — they know "deleted" isn't "gone."
OpenAI's argument is that their 30-day deletion window is reasonable and standard. Forcing them to stop that process mid-stream would create a massive backlog of data that their systems aren't designed to store indefinitely. It's not just about disk space. It's about database indexing, access controls, and the entire data lifecycle management architecture.
Think of it like a conveyor belt in a factory. Products move along, get processed, and exit. If you suddenly demand that every item be pulled off the belt and stored in a warehouse, the whole line backs up. Nothing moves. That's OpenAI's position — the deletion pipeline isn't a filing cabinet you can just lock. It's a process.
What This Means for ChatGPT Users
If you use ChatGPT, here's what you need to know right now.
Your deleted conversations are still being deleted. OpenAI hasn't changed its deletion policy. The court order hasn't taken effect yet — it's being challenged. So for now, when you delete a chat, it follows the normal 30-day removal process.
But this case highlights something important. Deleted doesn't always mean gone. And in the context of AI tools, your conversations are training data, legal evidence, and a privacy risk all rolled into one. I've started treating every AI conversation as potentially permanent, even when I hit delete. It's a healthy habit.
There's also a broader precedent here. If courts can force AI companies to preserve user data for litigation, it changes the privacy equation for every AI tool — not just ChatGPT. Claude, Gemini, Copilot — they'd all face the same pressure. This case could set the standard.
4 Privacy Practices for AI Tool Users in 2025
Based on what I've seen across multiple AI platforms, here are four things you should do now.
1. Assume nothing is truly private. Even with deletion policies, your conversations exist on servers for some period. If you're discussing something sensitive — trade secrets, personal health information, legal strategies — consider whether an AI tool is the right place for that conversation.
2. Check your AI tool's data retention policy. Not all platforms handle deletion the same way. Some retain data for training unless you opt out. Some delete immediately. Some keep logs for 30 days like OpenAI. Read the policy. It's boring but necessary.
3. Use the privacy features available. ChatGPT offers a "temporary chat" mode that doesn't save conversations to history. It's not perfect, but it's better than relying on deletion after the fact. Other tools have similar options.
4. Advocate for better defaults. The reason this court order is even possible is that AI companies collect and retain massive amounts of conversation data by default. Push for privacy-first defaults. Opt-in data collection, not opt-out.
I'll be honest — I don't follow all of these perfectly. I sometimes share things with AI tools that I probably shouldn't. Most of us do. But awareness is the first step toward better habits.
The Bigger Picture: AI Litigation and Your Data
This OpenAI case isn't happening in isolation. There are dozens of copyright lawsuits against AI companies right now. The New York Times is suing OpenAI and Microsoft. Getty Images sued Stability AI. Universal Music Group went after Anthropic. Every one of these cases will likely involve discovery requests for training data and user interactions.
According to a Reuters report on the case, the judge hasn't ruled yet on OpenAI's opposition. But the direction of these rulings matters enormously. If courts consistently side with plaintiffs on data preservation, AI companies will face a choice: fundamentally redesign their systems for legal compliance, or stop offering deletion features altogether.
Neither option is great for users. The first increases costs and complexity. The second eliminates the illusion of privacy entirely.
There's also a regulatory angle. The EU's AI Act and various proposed US regulations include data governance requirements. If courts are simultaneously demanding broad data preservation, AI companies get caught between conflicting obligations — delete data to comply with privacy laws, or keep it to comply with court orders.
This tension isn't going away. It's going to intensify as more lawsuits work through the system.
When I think about tools like AI-Mind, which generates content without requiring users to craft prompts, the privacy implications are slightly different but equally important. The less data you need to expose to get useful output, the smaller your privacy footprint. Zero-prompt tools that handle the engineering behind the scenes mean you're not iterating through dozens of prompt variations — each one a data point stored somewhere. It's not a perfect solution, but it's a step toward minimizing exposure. The first 30 generations are free, so you can test the approach without commitment.
Key Takeaways
- OpenAI is fighting a court order to preserve all ChatGPT conversations, including deleted chats, in a copyright lawsuit brought by authors.
- OpenAI argues the order is technically unworkable, overbroad, and undermines user privacy expectations.
- "Deleted" data often remains recoverable for days or weeks due to how cloud storage deletion pipelines work.
- This case could set a precedent for how AI companies handle user data in litigation across the entire industry.
- Users should treat AI conversations as potentially permanent and use privacy features like temporary chat modes when available.
Sources
- Reuters, OpenAI fights to preserve deleted ChatGPT logs in copyright case, March 2025. Coverage of OpenAI's court filing opposing the data preservation order.
- United States District Court, OpenAI Opposition Filing, March 2025. Legal document outlining OpenAI's arguments against the preservation order in Tremblay v. OpenAI.
- The Verge, Ongoing coverage of AI copyright lawsuits, 2024-2025. Context on the broader landscape of litigation against AI companies.
Frequently Asked Questions
Does OpenAI still delete my ChatGPT conversations when I click delete?
Yes. The court order has not taken effect — it's being challenged. OpenAI's current deletion process remains unchanged: deleted conversations enter a 30-day removal pipeline and are then permanently erased. This could change if the court rules against OpenAI, but for now, deletion works as advertised.
Can law enforcement or courts access my deleted ChatGPT history?
Potentially, yes. If a valid court order requires data preservation, companies must comply regardless of their normal deletion policies. This case shows that even deleted data can become subject to legal discovery. The outcome of this specific case will influence how broadly courts can demand AI conversation logs in future litigation.
How can I use ChatGPT more privately right now?
Use ChatGPT's "temporary chat" feature, which doesn't save conversations to your history. You can also disable chat history in settings, though this limits some features. For sensitive topics, consider whether an AI tool is appropriate at all — once data is shared, you lose control over how it's stored and potentially accessed.