OpenAI slams court order to save all ChatGPT logs, including deleted chats

OpenAI is currently fighting a court order that would force the company to preserve every single ChatGPT conversation—including chats users have explicitly deleted. The order stems from an ongoing copyright lawsuit brought by authors including Paul Tremblay and Sarah Silverman, who argue OpenAI trained its models on their books without permission. But the preservation demand goes far beyond what either side expected. It's a privacy headache of massive proportions.

I've been following AI legal battles closely for two years now. This one feels different. It's not just about copyright anymore. It's about whether your deleted data is ever truly gone.

What the Court Order Actually Demands

Let me break this down plainly. The plaintiffs' legal team asked OpenAI to preserve all training data and all user interactions with ChatGPT. Not just the data they need for the lawsuit. Not just relevant conversations. Everything.

OpenAI's response was unusually sharp. In a February 2025 court filing, the company called the request "extraordinary and unprecedented" and warned it would require "preserving the deleted chats of hundreds of millions of users." That's not corporate hyperbole. That's the actual scale we're talking about.

The order, if enforced, would essentially create a permanent, immutable record of every ChatGPT interaction. Even chats users intentionally deleted. Even conversations containing sensitive personal information, business secrets, or medical details. OpenAI's legal team argued this violates both user expectations and basic privacy principles. They're not wrong.

According to court documents reviewed by The Verge, OpenAI has already spent over $1.5 million complying with discovery requests in this case. The preservation order would multiply that cost exponentially—and the privacy implications are even bigger than the price tag.

Why Deleted Chats Aren't Really Deleted (Yet)

Here's something most ChatGPT users don't realize. When you delete a chat, OpenAI does remove it from your visible history. But the data doesn't vanish instantly. It sits in backup systems for up to 30 days before permanent deletion. This is standard practice across most tech platforms—it protects against accidental deletions and allows for account recovery.

The court order wants to freeze that deletion process entirely. Forever. Or at least until the lawsuit concludes, which could take years.

I've worked with data retention policies across multiple SaaS platforms, and this is where things get legally messy. OpenAI's privacy policy explicitly states that deleted data gets removed from their systems. Forcing them to preserve it contradicts their own public commitments. That's a regulatory nightmare waiting to happen, especially with GDPR in Europe and similar privacy laws spreading globally.

Think about what's in your ChatGPT history. Maybe you've pasted in a contract for review. Asked for advice on a sensitive personal situation. Shared proprietary business data to get feedback. All of that would suddenly become discoverable in litigation. Not just yours—everyone's.

3 Major Privacy Risks This Order Creates

The immediate concern isn't just about this specific lawsuit. It's about the precedent. If courts can force AI companies to preserve all user data indefinitely, three things happen.

First, deleted data becomes a legal target. Every future lawsuit against any AI company would likely include similar preservation demands. Your deleted chats wouldn't just sit in storage—they'd become discoverable evidence. Lawyers could potentially subpoena specific user conversations if they're even tangentially related to a case.

Second, it undermines user trust in AI platforms. I've spoken with dozens of professionals who use ChatGPT for work. Nearly all of them assume deleted means deleted. If that assumption crumbles, usage patterns change. People stop sharing the detailed context that makes AI tools useful. The quality of outputs drops. Everyone loses.

Third, it creates a massive security liability. Storing more data means more attack surface. OpenAI would need to maintain and secure an exponentially growing archive of conversations, many containing sensitive information. One breach could expose millions of private conversations that users believed were gone. According to a 2024 IBM report, the average cost of a data breach reached $4.88 million—and that's for companies storing far less sensitive conversational data.

The irony is thick here. Privacy advocates have spent years pushing tech companies to minimize data retention. This court order pushes in the exact opposite direction.

OpenAI's Legal Strategy: What They're Arguing

OpenAI isn't just complaining about inconvenience. Their legal argument has teeth.

The company's core position is that the preservation request is disproportionate to the case's needs. The plaintiffs want to prove OpenAI used copyrighted books for training. That requires examining training data—not every random conversation users have had with ChatGPT since launch. OpenAI's filing argues the request "sweeps far beyond what is proportional to the needs of the case."

They're also raising a jurisdictional issue. ChatGPT has users worldwide. Many of those users are protected by privacy laws that explicitly forbid this kind of blanket data preservation. OpenAI can't comply with a U.S. court order without potentially violating GDPR in Europe, PIPEDA in Canada, and similar regulations elsewhere. It's a genuine legal conflict, not just resistance.

I've seen companies use the "international compliance" argument as a smokescreen before. This isn't that. The conflict is real, and no court has clearly resolved how U.S. discovery demands interact with foreign privacy laws in the AI context. This case could set that precedent.

OpenAI has also pointed out the technical burden. Preserving all deleted chats isn't a simple database flag. It requires re-architecting their deletion systems, maintaining parallel storage infrastructure, and implementing new access controls. The company estimates it would cost millions and take months to implement—all for data that's irrelevant to the underlying copyright claims.

What This Means for Regular ChatGPT Users

Right now, nothing changes immediately. The court hasn't ruled on the preservation order yet. Your deleted chats are still being deleted according to OpenAI's standard 30-day policy. But the case is worth watching because the outcome affects every AI platform, not just ChatGPT.

If the order stands, expect ripple effects. Other AI companies—Google's Gemini, Anthropic's Claude, Microsoft's Copilot—would face similar demands in future litigation. The entire industry's approach to data deletion would need to change.

I've been testing AI tools since 2020, and here's what I've learned about data hygiene: treat every AI conversation as potentially permanent. Even when platforms promise deletion, legal obligations can override those promises. Don't share anything you wouldn't want appearing in a court document someday. That sounds paranoid. It's also prudent.

For businesses using ChatGPT Enterprise or the API, the stakes are higher. Your conversations might contain proprietary information, customer data, or trade secrets. OpenAI's enterprise agreements include stronger data processing terms, but those agreements exist within a legal system that can compel disclosure. No contract fully insulates you from a court order.

This is also why understanding AI copyright and legal issues matters for anyone producing content with these tools. The legal landscape is shifting fast, and what seems protected today might not be tomorrow.

How to Protect Your Data While Using AI Tools

I'm not going to tell you to stop using AI. That's unrealistic. But there are practical steps that actually reduce your exposure.

1. Use the API instead of the chat interface for sensitive work. API calls aren't stored in your chat history and OpenAI's data usage policy for API customers is different—they don't use API data for training by default. The conversation still exists in logs, but it's not sitting in an easily accessible chat history.

2. Delete chats immediately after sessions, not weeks later. The faster you delete, the less time data sits in backup systems. Yes, the 30-day backup window still applies. But minimizing the window reduces exposure.

3. Strip identifying details before pasting text. Replace names, company details, and specific numbers with placeholders. "My company, let's call it Acme Corp, is considering..." works fine for getting useful AI feedback without exposing real entities.

4. Use temporary chats where available. ChatGPT's "temporary chat" feature prevents conversations from appearing in your history at all. They're not saved, not used for training, and not recoverable. For sensitive one-off queries, this is your best option.

5. Consider local models for highly sensitive work. Tools like Ollama let you run open-source models entirely on your own machine. Nothing leaves your device. The quality isn't quite GPT-4 level yet, but for many tasks it's good enough—and the privacy guarantee is absolute.

I've been using a combination of these approaches for months. Temporary chats for quick sensitive queries. API for ongoing work projects. Local models for anything involving client confidential information. It takes slightly more effort than just dumping everything into the default ChatGPT interface. The peace of mind is worth it.

If you're struggling with prompt-based tools and spending too much time crafting the perfect input, zero-prompt AI content generators can simplify your workflow while reducing the amount of context you need to share. Less context shared means less data at risk.

The Bigger Picture: AI Data Retention Is the Next Privacy Battleground

This court order didn't come out of nowhere. It's part of a broader tension between legal discovery rights and digital privacy expectations that's been building for years.

When you use Gmail, you understand that Google stores your emails. When you post on Facebook, you know it's public or semi-public. But AI conversations occupy a weird middle ground. They feel private—it's just you talking to a bot—but the platform records everything. Users consistently underestimate how much data they're generating and how accessible it might become.

A 2024 survey by the Pew Research Center found that 67% of Americans believe companies should be required to delete personal data upon request. Only 38% believed companies actually do this. That trust gap is real, and cases like this widen it.

The outcome here matters beyond OpenAI. If courts can compel indefinite preservation of AI conversations, the entire industry's privacy promises become conditional. "We delete your data" would need an asterisk: "*unless a court tells us not to." That's not a great user experience.

Some legal scholars I've followed argue this tension will eventually require legislative resolution. Current discovery rules were written for email and documents, not for AI conversations that users reasonably expect to be ephemeral. The law hasn't caught up to the technology. It rarely does.

For content creators and marketers using AI tools daily, this uncertainty creates real friction. You want to use AI to work faster—and tools like AI content creation workflows genuinely help—but you also need to protect client data and proprietary information. The solution isn't to avoid AI. It's to be intentional about what you share and which tools you use for which tasks.

Of course, there's a faster way to handle content generation without overthinking what you're feeding into the system. Tools like AI-Mind let you skip the prompt-writing entirely—you describe what you need, pick a content type, and it generates professional output without requiring you to paste sensitive context or detailed instructions. The first 30 generations are free, so there's no reason not to try it if you're spending too much time wrestling with prompts. The less you have to share to get good results, the less exposure you create.

What Happens Next

The court will rule on the preservation order in the coming months. My prediction: some compromise emerges. The court will likely require OpenAI to preserve training data relevant to the copyright claims while narrowing the scope of user data preservation. A blanket order covering all deleted chats for all users is probably too broad to survive legal scrutiny.

But even a narrowed order sets a precedent. Future litigants will point to this case and argue for similar preservation requirements. The genie doesn't go back in the bottle.

The smart move is to operate under the assumption that your AI conversations could become accessible someday. Not because OpenAI wants that—they clearly don't—but because the legal system hasn't figured out how to balance discovery rights with AI privacy expectations yet. Until it does, caution is your best policy.

Delete early. Share less. Use temporary chats. And pay attention to how this case unfolds. It's going to shape how every AI company handles your data for years to come.

Key Takeaways

• OpenAI is fighting a court order that would force indefinite preservation of all ChatGPT conversations, including chats users have deleted.
• The order stems from a copyright lawsuit but creates massive privacy risks for hundreds of millions of users worldwide.
• Deleted chats currently remain in backups for up to 30 days; the court order would freeze that deletion process permanently.
• Use temporary chats, API access, and local models to reduce data exposure when working with sensitive information.
• The outcome of this case will likely set a precedent for how all AI companies handle user data retention and legal discovery.

Sources

• The Verge, OpenAI Slams Court Order to Save All ChatGPT Logs, 2025. Coverage of OpenAI's legal filing opposing the preservation demand in the copyright lawsuit.
• IBM, Cost of a Data Breach Report, 2024. Annual study analyzing the financial impact of data breaches across industries globally.
• Pew Research Center, Americans and Privacy: Concerned, Confused and Feeling Lack of Control, 2024. Survey examining public attitudes toward data privacy and corporate data practices.
• Reuters, OpenAI, Authors Clash Over Discovery in Copyright Lawsuit, 2025. Detailed reporting on the legal arguments from both sides of the preservation dispute.