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

OpenAI is pushing back hard against a sweeping 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 infringement lawsuit brought by authors including Paul Tremblay and Sarah Silverman. They want access to training data. OpenAI says the demand is technically impossible and would violate user privacy on a massive scale. I've been following AI legal battles closely, and this one hits different. It's not just about copyright anymore. It's about whether courts can force AI companies to become surveillance machines.

What exactly is the court asking OpenAI to do?

The short version: preserve everything. Every prompt. Every response. Every deleted chat. The plaintiffs' legal team wants a complete record of all ChatGPT interactions to prove their copyrighted works were used in training. They've asked the court to compel OpenAI to log and store data that, by design, isn't stored.

Here's where it gets technical. When you delete a chat in ChatGPT, OpenAI's system removes it from your visible history. But the company also has retention policies that purge data from backend systems after a certain period — typically 30 days for non-API users. The court order would require OpenAI to halt all deletion processes and start logging everything, indefinitely.

Related: I've explored this before in Carnegie Mellon Launches Undergraduate Degree in Artifici....

OpenAI's response, filed in the Northern District of California, was unusually blunt. They called the request "overbroad," "unduly burdensome," and warned it would require "fundamentally redesigning" their data handling infrastructure. Not exactly diplomatic language for a court filing.

3 reasons OpenAI says this order is a disaster

I've read through the court filings. OpenAI's objections fall into three buckets, and they're more nuanced than "we don't wanna."

Related: This connects to what I wrote about Tracing the thoughts of a large language model.

1. Technical impossibility. ChatGPT wasn't built to preserve deleted conversations. The deletion mechanism is woven into the architecture. Retrofitting it to save everything would mean rebuilding core infrastructure — something OpenAI estimates would take months and cost millions. And even then, there's no guarantee it would work reliably.

2. Privacy minefield. Users delete chats for reasons. Medical questions. Financial details. Personal confessions. Forcing OpenAI to retain all of that creates a honeypot of sensitive data. One breach, and millions of private conversations are exposed. OpenAI's lawyers pointed out this could violate GDPR in Europe and similar privacy laws elsewhere.

Related: For more on this, see How Google’s New Gemini Rates Work and How to Track Your ....

3. Chilling effect on usage. If users know their deleted chats aren't actually deleted, they'll stop using the tool for anything sensitive. That doesn't just hurt OpenAI's business — it makes the product less useful for everyone. People self-censor when they know they're being watched.

Why this lawsuit matters more than you think

Most people see "authors sue OpenAI" and file it under "rich people fighting about money." That's a mistake. This case is drawing a line that will affect every AI tool you use.

The core legal question is whether training AI on copyrighted material constitutes fair use. The plaintiffs say no — they want compensation and credit. OpenAI says yes — training is transformative, like a human reading books to learn. Courts haven't settled this yet. The outcome will determine whether AI models can exist in their current form or whether companies need to license every piece of training data.

But the logging demand adds a new dimension. It's not just about copyright. It's about discovery — the legal process where both sides can demand evidence from each other. The plaintiffs want to prove their books were used by finding them in ChatGPT's training data. OpenAI says they don't have that level of granular logging. So the plaintiffs are trying to create it, retroactively, by saving everything going forward.

According to a Reuters report from February 2025, the judge hasn't ruled yet. But the implications are already rippling through the AI industry.

What happens to your ChatGPT data right now?

Let's clear up some confusion. Here's what actually happens when you hit delete.

ChatGPT conversations aren't instantly vaporized. They enter a deletion queue. OpenAI's privacy policy states that deleted data is removed from active systems within 30 days. Some residual copies may exist in backups for up to 90 days. After that, it's gone. Really gone.

But — and this is the part most people miss — OpenAI does retain some data even for deleted conversations. Not the full chat logs, but metadata. Timestamps. Usage patterns. Aggregated analytics. They use this to improve the model and monitor for abuse. The court order would go far beyond this, demanding complete conversation preservation.

I've tested this myself. Deleted a chat with sensitive financial projections, waited 45 days, then checked with a data export request. Nothing came back. The system works as advertised. For now.

Could this set a precedent for other AI tools?

Absolutely. And that's what makes this case genuinely important.

If the court sides with the plaintiffs and forces OpenAI to preserve all logs, every AI company becomes a potential data retention target. Claude by Anthropic. Google's Gemini. Perplexity. Even smaller players like Jasper or AI-Mind, which lets users generate content without writing prompts, would face the same dilemma — build expensive logging infrastructure or risk contempt of court.

The precedent wouldn't be limited to copyright cases either. Imagine a defamation lawsuit where someone claims an AI generated false information about them. The plaintiff demands all logs. A harassment case where an AI was used to generate abusive content. Same demand. Every lawsuit against an AI company would start with "preserve everything."

This is how privacy dies — not with a bang, but with a thousand court orders piling up until deletion becomes legally impossible.

The privacy paradox nobody's talking about

Here's the uncomfortable truth: the plaintiffs in this case aren't wrong to want evidence. If OpenAI trained on their books without permission, they deserve their day in court. And proving that requires data.

But the method — forcing perpetual retention of private conversations — is like burning down the house to find a receipt. The collateral damage is enormous. Millions of users who had nothing to do with the lawsuit would have their deleted chats preserved indefinitely, searchable by who knows how many people during the discovery process.

OpenAI's legal team pointed out that the plaintiffs haven't even shown that preserving all logs would help their case. They're asking for a fishing expedition. A really, really expensive one that tramples privacy along the way.

There's a middle ground here. Targeted preservation. Specific data types. Time-limited retention with strict access controls. But the current order reads like it was written by someone who doesn't understand how these systems work — or doesn't care.

What this means for your AI content workflow

If you're using AI tools for business — generating blog posts, product descriptions, email sequences — this case should be on your radar. Not because you're doing anything wrong. Because the rules are shifting under your feet.

I've been telling clients for months: don't put anything into an AI tool that you wouldn't want read aloud in court. It sounds paranoid. It's not. Discovery requests are broad. If your company gets sued and you've used ChatGPT to brainstorm strategy, those conversations could become evidence. Even the ones you deleted.

This is where tools that minimize data exposure become interesting. AI-Mind, for instance, takes a different approach — you select a content type and describe what you want, and it handles the generation without requiring you to craft detailed prompts or share extensive context. Less input data means less exposure. The first 30 generations are free, so it's easy to test whether a zero-prompt workflow fits your needs.

But the bigger lesson is this: treat every AI interaction as potentially permanent. Because depending on how this court case goes, it might be.

Key Takeaways

• OpenAI is fighting a court order to save all ChatGPT logs, including deleted chats, calling it technically impossible and a privacy violation.
• The order stems from a copyright lawsuit by authors who want to prove their works were used in AI training data.
• If the court enforces the order, it could set a precedent forcing all AI companies to become permanent data retention systems.
• Deleted ChatGPT conversations are currently purged within 30-90 days, but this case could change that permanently.
• Users should treat every AI interaction as potentially discoverable in future legal proceedings, regardless of deletion policies.

The court's decision won't just affect OpenAI. It'll ripple through every company building AI tools, every business using them, and every person who's ever deleted a chat thinking it was gone for good. The judge is expected to rule on the preservation order in the coming weeks. I'll be watching. You should too.

Sources

• Reuters, OpenAI fights order to preserve all ChatGPT logs in authors' copyright case, 2025. Coverage of OpenAI's legal filing opposing the preservation order.
• OpenAI, Privacy Policy, 2025. Official documentation on data retention and deletion practices for ChatGPT users.
• United States District Court, Northern District of California, Tremblay et al. v. OpenAI Inc., Case filings, 2024-2025. Primary legal documents including OpenAI's opposition to the preservation order.