Is ChatGPT Safe for Company Documents? A 2026 Guide

Whether ChatGPT is safe for company documents depends mostly on which ChatGPT your team uses. On the free and personal plans, OpenAI may use your conversations to train its models unless you turn that off. On ChatGPT Business, Enterprise and Edu, it doesn’t train on your data by default. But training is only one of the risks. Where your data is stored, how long it’s kept and whether the answers are correct matter just as much.

Your team is probably already doing it. In KPMG’s 2025 study with the University of Melbourne, 46% of US workers admitted to uploading sensitive company information and intellectual property to public AI platforms. Globally, 57% said they hide their AI use at work.

This guide explains what each ChatGPT plan does with your data, the five risks to check, and when a dedicated AI knowledge base is the better choice.

This article is general information, not legal advice. It reflects OpenAI’s published policies as checked on September 30, 2026. Policies change, so always check OpenAI’s current terms.

Key Takeaways

• Free, Plus and Pro: OpenAI may use your conversations to train its models unless you switch off “Improve the model for everyone.” Don’t put confidential company documents into personal accounts.
• Business, Enterprise and Edu: no training on your data by default, SOC 2, and a data processing agreement is available. Only Enterprise and Edu let admins control retention.
• “Deleted” isn’t always gone. A 2025 US court order made OpenAI preserve consumer and Team chats from May to September 2025, but not Enterprise or Edu data.
• Even a safe setup can give wrong answers. For company knowledge, use a tool that answers from your own documents and shows the source.

Table of contents

What does ChatGPT do with your company data?

It depends on the plan. Here’s what OpenAI’s own help center and policy pages say:

Personal plans. OpenAI’s help center says that for individual services such as ChatGPT, OpenAI may use your content to train its models. To opt out, turn off “Improve the model for everyone” under Settings > Data controls. One detail most people miss: even after opting out, giving a thumbs up or down means the entire conversation may be used for training. Temporary Chats, on the other hand, aren’t used to improve the models while they stay temporary.

Business plans. For ChatGPT Business, Enterprise, Edu and the API, OpenAI states that by default it doesn’t use inputs or outputs to improve its models. Its enterprise privacy page also mentions a completed SOC 2 audit, encryption at rest (AES-256), and a data processing agreement for Business, Enterprise and API customers to support GDPR compliance. Retention control is listed for Enterprise, Edu and Healthcare, not for Business.

Why employees use ChatGPT for company documents anyway

Because it works, and because using it is easier than asking permission. The KPMG and University of Melbourne study, which surveyed more than 48,000 people in 47 countries, found widespread use of AI outside company policy. It also found that many people trust the output a little too much.

The risk is real. In 2023, Samsung banned generative AI tools on company devices after engineers uploaded internal source code to ChatGPT, according to a memo reviewed by Bloomberg.

Five risks to check before you put company documents into ChatGPT

1. Training on your content

On personal plans, training is on by default. If employees use personal accounts for work, your documents may end up as training data unless each person has opted out. That’s why the plan matters more than any single setting.

2. Retention and legal holds

Deleted chats are normally removed within 30 days. But a court can change that. In May 2025, a US court in The New York Times v. OpenAI ordered OpenAI to preserve output logs, including chats users had deleted. According to OpenAI’s own update, which says it contested the order, this affected Free, Plus, Pro and Team (now Business) users and API customers without zero data retention, but not ChatGPT Enterprise or Edu. The obligation ended on September 26, 2025, but OpenAI says it still stores limited historical data from April to September 2025 under legal hold. The lesson is that “deleted” doesn’t always mean gone.

3. Where your data is stored

For EU and UK companies, location matters. OpenAI’s help center says data residency for ChatGPT Business is “rolling out gradually and is not yet available to all customers”. Even with a non-US region selected, “a copy of every prompt and response is also stored in the United States for a limited time”, abuse-monitoring logs stay in the US, and residency doesn’t cover where requests are processed.

4. Who can see what

A tool that can read all your documents can also surface them to the wrong person. ChatGPT’s Company Knowledge feature, available on Business, Enterprise and Edu, respects permissions in the connected source, so members only retrieve what they’re already allowed to access. Check your source permissions first: if a salary sheet is shared with everyone in Google Drive, an AI assistant will treat it that way too.

5. Whether the answers are right

Privacy is only half the question. OpenAI’s own researchers wrote in 2025 that language models sometimes guess when uncertain, producing plausible yet incorrect statements. In Mata v. Avianca (2023), a US court fined lawyers $5,000 for submitting a brief with case citations ChatGPT had made up. For company policies, a confident wrong answer about leave or expenses can cause real problems. Our guide to the best AI knowledge base tools explains why visible sources matter.

What about GDPR and regulators?

GDPR doesn’t give a simple yes or no for ChatGPT. What matters is how your company uses it: which plan, which data, whether you have a data processing agreement, and whether you’ve told employees and customers how their data is handled.

Regulators are watching, but the outcomes vary. In December 2024, Italy’s data protection authority fined OpenAI €15 million, among other things for training on personal data without first identifying an adequate legal basis. In March 2026, a court in Rome annulled that fine, Reuters reported. For your company, the practical takeaway is the same: use a business plan with a DPA, and keep personal data out unless you’ve assessed it.

How to use ChatGPT safely with company documents

If your team uses ChatGPT, set clear rules rather than hoping for the best:

1. Choose a business plan. Use ChatGPT Business or Enterprise for work, not personal accounts, and put OpenAI’s DPA in place.
2. Write a short AI policy. Say which tools are approved and which data never goes in: personal data, customer data, contracts, financials and source code, unless the tool is approved for it.
3. Turn off training on any personal accounts that are still used, and remind people that feedback buttons can share a conversation.
4. Fix permissions before connecting sources. Review who can see what in Google Drive or SharePoint before any AI tool can search it.
5. Check where data is stored if you’re in the EU or UK, including the temporary US copy on Business.
6. Always verify answers against the original document before acting on them.

A ban alone rarely works. When more than half of employees in KPMG’s study say they hide their AI use, a blanket ban mostly pushes it further out of sight. Give people an approved tool instead, and put your AI policy in your employee handbook.

When is a dedicated AI knowledge base the better choice?

ChatGPT is a general-purpose assistant. For questions about your own company, such as policies, processes and “how we do things here”, a dedicated AI knowledge base is often the better fit:

A common setup is to use both: ChatGPT for writing and analysis, and a knowledge base as the single trusted home for company answers. If you’re building that home, our guide to creating an internal knowledge base walks you through it step by step.

Frequently asked questions

Is it safe to upload confidential documents to ChatGPT Plus if I turn off training?

It’s safer, but not ideal for company documents. Turning off training stops new chats from being used to improve the models, but chats are still stored until you delete them, deleted chats take up to 30 days to be removed, and legal holds can override that. For confidential company documents, use a business plan with a DPA, or a tool built for company knowledge.

Does ChatGPT Business train on our data?

No, not by default. OpenAI states that by default it doesn’t use inputs or outputs from ChatGPT Business, Enterprise, Edu or the API to improve its models.

Is ChatGPT GDPR compliant?

There’s no single yes or no. OpenAI offers a data processing agreement for Business, Enterprise and API customers, and it offers EU data residency for eligible plans. Whether your use is compliant depends on what data you put in, your legal basis, and your agreements. Check with your data protection officer or lawyer.

Should we ban ChatGPT at work?

A ban can make sense for highly sensitive environments, as Samsung decided in 2023. For most companies, a clear policy plus an approved business tool works better, because bans tend to push AI use out of sight.

Safe enough, if you choose the setup on purpose

• Personal plans may train on your data by default. Keep company documents out.
• Business plans don’t train by default and offer a DPA, but check retention and where data is stored.
• Legal holds can keep “deleted” data longer than you expect.
• Permissions and accuracy matter as much as privacy.
• For company knowledge, a dedicated AI knowledge base that answers from your own documents with sources is often the safer choice, and easier to verify. It also helps when employees keep asking the same questions.