Skip to main content
Get your AI Readiness score in 5 minutes— free, no signup, no salesperson will call.
Freistyle AI

AI Education

Is ChatGPT Private? What the 2026 Court Rulings Mean for Your Business

Your ChatGPT chats can end up in court, and some plans use them for training. What the rulings say, how the plans differ, and a one-hour fix for a small business.

By Dominic FreiPublished Updated 12 min read

Not in the way most people assume. ChatGPT chats aren't legally confidential: a court can order them to be preserved and handed over, and a 2026 ruling found that chatting with an AI isn't protected by attorney-client privilege. Which plan you use also decides whether OpenAI can train on your chats. For a business, that changes what you should type into it.

I wrote the first version of this in May. Since then the Heppner ruling has been analyzed in detail, the EU moved one of its AI deadlines, and OpenAI renamed and changed several of the settings I described. So I've rewritten it. One note before you read on: I'm not a lawyer. I spent 15 years in corporate risk and security, and this is general information. For a specific legal matter, talk to your lawyer.

Is ChatGPT private when you use it for work?

It depends on what you mean by private. Three different ideas get mixed up:

  • Encryption: for its business products, OpenAI says it encrypts data with AES-256 at rest and TLS 1.2 or higher in transit (OpenAI). This protects your chats from outsiders.
  • Confidentiality: OpenAI can access conversations in some cases, for example when a small internal team reviews chats flagged for possible harm to others. If the reviewers find an imminent threat of serious physical harm to others, OpenAI may refer the case to law enforcement (OpenAI, August 2025).
  • Legal privilege: this decides whether a court can force your chats to be handed over. For AI chats, the answer so far has been yes.

Sam Altman, OpenAI's CEO, said in July 2025 that the confidentiality people get with a doctor or lawyer doesn't exist yet for ChatGPT: "we haven't figured that out yet for when you talk to ChatGPT" (TechCrunch, July 2025). For a business owner, privilege is the one that matters most.

Can your ChatGPT chats be used in court?

Yes. Two cases show how.

The first is United States v. Heppner in the Southern District of New York. A man facing fraud charges had used Claude, Anthropic's chatbot, to prepare material about his defense before he was indicted, without his lawyers asking him to. The FBI seized the documents. Judge Jed Rakoff ruled on February 10, 2026, with a written opinion on February 17, that the chats were not protected by attorney-client privilege or as work product (Harvard Law Review, March 2026). His reasons are worth knowing:

  • Claude is not a lawyer, so there was no attorney-client relationship.
  • The chats weren't confidential, because Anthropic's privacy policy allows it to collect conversations and disclose them to third parties, including in litigation.
  • He didn't use the AI at his lawyers' direction. The judge suggested the answer might have been different if counsel had directed it, and the ruling doesn't address enterprise AI products (Proskauer, 2026).

The ruling was about Claude, but nothing in the reasoning is specific to Anthropic, so I'd assume a court would see ChatGPT chats the same way.

The second is the New York Times copyright case against OpenAI. In May 2025 a court order required OpenAI to keep consumer ChatGPT chats, including ones users had deleted. The order covered ChatGPT Free, Plus, Pro and Team (now Business), and the API without a zero data retention agreement. It didn't cover ChatGPT Enterprise, Edu or zero data retention API customers. The obligation ended on September 26, 2025, but OpenAI says it still holds a limited set of April to September 2025 data separately, accessible only to a small legal and security team (OpenAI, updated October 2025). Separately, a magistrate judge ordered OpenAI to produce a sample of 20 million de-identified ChatGPT logs under a protective order, and District Judge Sidney Stein affirmed that order on January 5, 2026 (ABA Journal, January 2026).

Criminal cases use chat logs too. When federal prosecutors charged a man with starting the fire that became the Palisades Fire in Los Angeles, they cited his ChatGPT history, including an image he had generated of a burning forest (ABC7, October 2025).

For most small businesses, the realistic risk is civil: a dispute with a former employee, a customer, a partner or a regulator. In US civil cases, each side can ask for any non-privileged information relevant to the claims, within limits of proportionality, under Federal Rule of Civil Procedure 26 (Cornell LII). If someone at your company used ChatGPT on the matter, those chats can be part of that.

Does deleting a ChatGPT chat delete it?

Usually, within 30 days. When you delete a chat, OpenAI schedules it for permanent deletion within 30 days, unless it was already de-identified and separated from your account, or OpenAI has to keep it longer for security or legal reasons (OpenAI Help). The 2025 court order showed what that exception can mean in practice.

Memory is separate. Deleting a chat doesn't necessarily delete a saved memory that ChatGPT created from it, so check Settings and delete the memory as well. OpenAI may keep logs of deleted memories for up to 30 days (OpenAI Help).

Temporary Chats are deleted within 30 days and aren't used for training (OpenAI Help). Before you start one, you can choose whether it uses your existing memories. Pick "Unpersonalized" if you don't want it to.

Which ChatGPT plans train on your data?

The individual plans do, unless you switch it off. OpenAI says it may use content from Free, Plus and Pro to train its models. To stop that, turn off "Improve the model for everyone" under Settings, then Data controls. By default, OpenAI doesn't use inputs or outputs from ChatGPT Business, Enterprise, Edu or the API for training (OpenAI Help). Switching training off only affects new conversations, and it doesn't change what a court can order.

PlanUsed for training by default?Covered by the 2025 NYT preservation order?What you get for business use
Free, Plus, ProYes, unless you switch it offYesConsumer terms only
Business (formerly Team)NoYesWorkspace with admin controls; a Data Processing Addendum is available
Enterprise and EduNoNoRetention controls, Data Processing Addendum, and for eligible customers your own encryption keys and data residency
API with zero data retentionNoNoFor qualifying organizations; prompts aren't stored

ChatGPT Business costs $25 per user a month billed monthly, or $20 billed annually (OpenAI Help). On annual billing that's the same as a Plus subscription, and your chats are no longer used for training. In my view, that's the cheapest privacy upgrade a small team can make.

Two limits are worth knowing. Data residency, which means storing your data in a region such as Europe, is offered for eligible Enterprise, Edu, Healthcare and API customers, and Business isn't listed. A HIPAA Business Associate Agreement is only offered for ChatGPT for Healthcare and API healthcare customers (OpenAI).

What should a business never put into ChatGPT?

Anything you wouldn't want read aloud in a deposition or sent to a competitor. My hard "never" list:

  • Customer personal data, payment details, passwords and API keys
  • Source code for login, payments or anything proprietary
  • Contracts, NDAs, term sheets and deal documents
  • Personnel files, salaries and HR investigations
  • Patient or medical information
  • Legal strategy and anything your lawyer told you in confidence

Then there's a "redact first" list. Financial figures, supplier names, pricing strategy and internal disputes can go in once you've swapped the real details for placeholders like "Client A", "Supplier B" and round numbers. It takes a minute and removes most of the risk.

Samsung learned this early. In 2023 it restricted generative AI on company computers after staff uploaded sensitive code to ChatGPT (CNBC, May 2023).

Are the rules stricter in Switzerland and the EU?

In some ways, yes. Switzerland's revised data protection law (nFADP) has applied since September 1, 2023, and the Federal Data Protection and Information Commissioner (FDPIC) has said it applies directly to AI. Providers and users have to be transparent about the purpose of AI-based processing, people have a right to know when they're talking to a machine, and high-risk processing needs a data protection impact assessment (FDPIC, updated May 2025). If you serve EU customers, the GDPR applies as well.

The EU AI Act timeline changed this summer. Regulation (EU) 2026/1744 entered into force on July 27, 2026 and moved the obligations for stand-alone high-risk AI systems, such as AI used in hiring or credit decisions, from August 2, 2026 to December 2, 2027. Most of the transparency rules in Article 50 still apply from August 2, 2026, and the duty to promote AI literacy among staff stays, in a softer form (White & Case, August 2026).

What does this mean if your business ends up in a dispute?

Treat AI chats like email. When a lawsuit is reasonably likely, you generally have to preserve relevant records, and chats about the matter in ChatGPT or any other AI tool can be among them. Three rules belong in your employee handbook now:

  1. Company work happens only in the company's ChatGPT Business or Enterprise workspace, not in personal accounts.
  2. Some categories of information never go into any AI tool.
  3. Once a legal hold is announced, nobody deletes chats about the matter.

If you work with lawyers, read how they handle this side. My guide to how lawyers are using AI covers the ABA's ethics opinion on AI and the fake-citation cases.

A one-hour fix for a small business

You can cover most of the risk in about an hour:

  1. List who uses AI for work (10 minutes): name, tool, personal or company account, and what they use it for.
  2. Switch off training (10 minutes) on every personal Free, Plus or Pro account that's still used for work.
  3. Pick the right plan (15 minutes): ChatGPT Business at minimum for client data. Look at Enterprise or the API with zero data retention if you're in a regulated field or handle trade secrets.
  4. Write a one-page AI use policy (15 minutes): approved tools, the never list, the redact-first list, what to do after a mistake, and the legal hold rule.
  5. Tell the team (10 minutes) in a short meeting or video, and get a written acknowledgment from each person.

For step 4, I've since written the policy out in full, with a copy-ready template and the EU rules that apply, in how to write a one-page AI policy for your business.

Quick Win
Open the ChatGPT account each person uses for work and check three things: is it a personal account or a company workspace, is "Improve the model for everyone" switched off, and what's saved in Memory? Put the answers in a simple spreadsheet. That sheet is the start of your AI policy.

When is ChatGPT safe enough to use?

For most everyday work, on a Business or Enterprise plan with a written policy. A simple traffic-light system helps:

UseExamplesMinimum setup
GreenMarketing copy from public information, summarizing public documents, general emails, brainstormingAny plan with training switched off
YellowFinancial models with anonymized figures, HR drafts without names, contract drafts with placeholdersBusiness or Enterprise, after redacting
RedRegulated data, privileged legal material, anything you couldn't defend in courtEnterprise or API with zero data retention, after legal advice, or don't use AI

Before you paste something, ask yourself whether you'd be comfortable seeing it read out in a deposition. If you wouldn't send it in an unencrypted email, don't paste it into a chatbot.

Related Tool
The free AI Readiness Quiz asks 16 questions about your business and scores your readiness across 5 areas: data, process, technology, team and strategy. It takes about 5 minutes and needs no signup.

Frequently asked questions

Is ChatGPT private?

Not legally. Chats aren't covered by attorney-client or similar privilege, and courts can order them to be preserved and produced. On Free, Plus and Pro, OpenAI may also use them for training unless you switch that off.

Can my ChatGPT conversations be used against me in court?

Yes. In United States v. Heppner, a federal judge ruled in February 2026 that a defendant's AI chatbot conversations weren't privileged. In the New York Times case against OpenAI, a judge affirmed an order to produce 20 million de-identified ChatGPT logs.

If I delete a ChatGPT chat, is it gone?

Usually within 30 days. OpenAI keeps deleted chats longer if it has a legal or security obligation, as it did under a court order in 2025. Saved memories are separate and have to be deleted on their own.

Does ChatGPT train on my conversations?

On Free, Plus and Pro it may, unless you turn off "Improve the model for everyone" in Settings under Data controls. ChatGPT Business, Enterprise, Edu and the API aren't used for training by default.

What's the difference between ChatGPT Plus and ChatGPT Business for privacy?

Business doesn't use your chats for training by default, adds a company workspace with admin controls, and offers a Data Processing Addendum. It costs $20 per user a month on annual billing, the same as Plus, or $25 billed monthly. Business was also covered by the 2025 preservation order, so it doesn't protect chats from court orders.

Is ChatGPT allowed under Swiss and EU data protection law?

It can be used, but the rules apply in full. Switzerland's data protection commissioner says the nFADP applies directly to AI, including transparency and impact assessments for high-risk processing. EU businesses also have to meet the GDPR. Check with your data protection adviser before putting personal data into any AI tool.

What's the fastest thing a business owner can do today?

Move work into a company workspace, switch off training on any personal account that's still used for work, and write down what must never be pasted into AI. That covers most of the risk in about an hour.

Where to go from here

Run the three-question check in the Quick Win box today, then block an hour this week for the full fix. If you have a lawyer, send them your one-page policy before you roll it out.

If you'd like one short, practical note on AI and risk a week, The Risk Memo goes out every Tuesday.

What changed in this update (October 27, 2026)
I checked every legal claim against the ruling summaries and OpenAI's own help pages, and corrected several. The plan table now shows that ChatGPT Business, then called Team, was fully covered by the 2025 preservation order. I also removed details about the Heppner case and the Samsung incident that I couldn't confirm, and corrected a claim about Federal Rule 26. The EU section now reflects the new December 2027 date for high-risk AI rules. New: the Palisades Fire case, OpenAI's police referral policy, memory and Temporary Chat settings, an FAQ and the full list of sources.

Researched with AI tools, written and fact-checked by Dominic.

Sources

  1. OpenAI, Business data privacy, security, and compliance
  2. OpenAI, Helping people when they need it most (Aug 26, 2025)
  3. TechCrunch, Sam Altman warns there's no legal confidentiality when using ChatGPT as a therapist (Jul 25, 2025)
  4. Harvard Law Review Blog, United States v. Heppner (Mar 23, 2026)
  5. Proskauer, SDNY Addresses Privilege and Work Product Implications of Using Unsecured Public AI Tools (2026)
  6. OpenAI, How we're responding to The New York Times' data demands in order to protect user privacy (Jun 5, 2025; updated Oct 22, 2025)
  7. ABA Journal, ChatGPT creator must turn over 20M chat logs in copyright litigation, federal judge says (Jan 8, 2026)
  8. ABC7 Los Angeles, 29-year-old man accused of starting blaze that turned into destructive Palisades Fire (Oct 8, 2025)
  9. Cornell Law School LII, Federal Rules of Civil Procedure, Rule 26
  10. OpenAI Help Center, Chat and file retention in ChatGPT
  11. OpenAI Help Center, Memory in ChatGPT
  12. OpenAI Help Center, Data controls in ChatGPT
  13. OpenAI Help Center, How your data is used to improve model performance
  14. OpenAI Help Center, ChatGPT Business overview
  15. CNBC, Samsung bans use of AI like ChatGPT for staff after misuse of chatbot (May 2, 2023)
  16. FDPIC (EDÖB), Update: Current data protection legislation is directly applicable to AI (May 8, 2025)
  17. White & Case, EU AI Omnibus enters into force, amending the AI Act (Aug 4, 2026)

Last updated: