Guide · AI and client data

Can I upload client files to ChatGPT?

Short answer: not as is, unless you know three things. Which ChatGPT plan you are on. What OpenAI does with what you send on that plan. And what your own professional rules say. Here is each one, in plain words, plus a safer way to do the work.

You have a contract, a tax file, a set of session notes or an HR complaint. ChatGPT could summarize it in a minute. Before you drag it in, read this. It is written for lawyers, accountants, therapists and HR staff. It quotes OpenAI's own help and privacy pages as they read on October 24, 2026. Those pages change, so check them again before you rely on them.

The short answer

The file goes to OpenAI. Whatever you upload or paste is sent to OpenAI's systems. No setting changes that. Settings change what OpenAI does with it next.
Your plan matters a lot. On Free, Plus and Pro, OpenAI may train on your chats unless you opt out. On Business and Enterprise, it does not by default.
Your rules matter more. A lawyer, a tax preparer and a therapist each have their own rule about sharing client information. A good plan does not answer those rules for you.
The safest version sends less. Take the names and numbers out first. Then the AI never gets them, on any plan.

What OpenAI says about Free, Plus and Pro

OpenAI's help page on model training says:

"When you use our services for individuals (Free, Plus, Pro), we may use your content to train our models."

You can opt out. The same page says to turn off Improve the model for everyone under Settings, then Data controls. After that, OpenAI says it won't use your new conversations to improve its models.

Three things the opt-out does not cover

Thumbs up and thumbs down. The page says that if you give feedback, "the entire conversation associated with that feedback may be used to train our models." That holds even after you opt out.
Saved chats stay saved. OpenAI's retention page says regular chats stay in your account until you delete them. A deleted chat is scheduled for permanent deletion within 30 days, unless OpenAI must keep it longer for security or legal reasons.
Temporary chats are kept for a while too. Temporary chats are not used for training while they stay temporary. But OpenAI says it "may keep a copy of a temporary chat for up to 30 days for safety purposes."

So opting out is worth doing. It is not the same as the file never leaving your hands.

What OpenAI says about Business and Enterprise

The same training page says:

"By default, we don't use inputs or outputs from ChatGPT Business, ChatGPT Enterprise, ChatGPT Edu, or our API to improve our models."

OpenAI's enterprise privacy page adds more for these plans:

Admins set retention. On Business, Enterprise and Edu, workspace admins control how long data is kept. Deleted conversations are removed within 30 days, unless the law or safety needs require longer.
Some people can still see chats. On Business, admins can view and export users' chats. OpenAI says its own access is limited to authorized employees for support, abuse checks and legal needs, and to contractors who review for abuse and misuse.
Health data is a separate question. OpenAI says it can sign a Business Associate Agreement for its API Platform, and it describes a separate product called ChatGPT for Healthcare. If HIPAA applies to you, read the therapists and HIPAA guide.

A business plan sends less of your data into training. It is still a third party holding your client's file.

What your own rules say

OpenAI's settings tell you what OpenAI does. They do not tell you what you are allowed to do. That depends on your work:

Lawyers. ABA Formal Opinion 512 says you need informed consent before putting client information into a self-learning AI tool. A line in your engagement letter is not enough. See the ABA Opinion 512 guide, and the California and New York guides.
Tax preparers and CPAs. Section 7216 of the tax code limits how you disclose or use a client's return information without written consent. See the IRC 7216 guide.
Therapists and counselors. HIPAA may apply, and some states now have rules about AI in therapy. See the therapists guide.
HR staff. There is no one rule for HR. Your company's own data policy decides most of it, along with any privacy or employment law where you work. Ask before you paste an employee's file.

A safer way: take the names out first

Most tasks do not need the real names. "Summarize this lease" works just as well when the tenant is [PERSON_1]. So send the AI a copy with the identifiers swapped out, and put them back in its answer afterward.

Check your plan and settings. Know whether training is on, and how long chats are kept.
Check your rules. Find out if your profession needs client consent first. If it does, get it.
Swap out the identifiers. Replace names, SSNs, account numbers, addresses, phone numbers and dates of birth with placeholders.
Read the result before you paste. Look for anything else that points to the client: a job title, an employer, a rare event.
Put the real values back in the AI's answer, on your own computer.

How to do it with KeptPDF

Anonymize for AI looks for names, emails, phone numbers, SSNs, account numbers, addresses, dates and many kinds of ID numbers. It swaps each one for a numbered placeholder like [PERSON_1], so the same person keeps the same label all through the file. It runs in your browser. The file is not uploaded to KeptPDF.

Open Anonymize for AI and drop in a PDF or Word file, or paste text.
Review what it found. Add anything it missed.
Copy the anonymized text into ChatGPT, Claude or whatever you use.
Put the real values back. Paste the AI's answer into the restore box. To restore later, click Save case key first. You can lock the key file with a passphrase.

Open Anonymize for AI

One honest note. The anonymized text you paste does go to the AI company. That is the point. Placeholders take out the direct identifiers, not every clue, so read the text before you paste it. And this step lowers what is exposed. It does not replace client consent where your rules require it. If a detail should reach no one at all, the Redact tool removes it from the PDF for good.

Sources

Quotes are from OpenAI's own pages as they read on October 24, 2026. Outside the short quotes, we summarize in our own words.

This is not legal advice. It is a plain-English summary for general information. Check your own professional rules, or ask your counsel, before you set an AI policy.

Try it: Anonymize for AI. Related reading: redact before pasting into ChatGPT, and do online PDF tools upload your file?

FAQ

Is it safe to upload client documents to ChatGPT?

It depends on your plan, your settings and your professional rules. Whatever you upload goes to OpenAI. On Free, Plus and Pro, OpenAI says it may train on your content unless you opt out. On Business and Enterprise, it says it does not train on your data by default. The safest version takes the names and numbers out first.

Does turning off "Improve the model for everyone" make it private?

It stops OpenAI from training on your new conversations. It does not stop the upload, and OpenAI says feedback you give with thumbs up or down can still put that whole conversation into training. Saved chats stay until you delete them.

Is ChatGPT Business or Enterprise safe for client files?

OpenAI says it does not train on Business or Enterprise inputs by default, and admins control retention. The file is still held by a third party, and your own rules on client consent still apply.

Are temporary chats deleted right away?

No. OpenAI says temporary chats are not used for training while they stay temporary, but it may keep a copy for up to 30 days for safety purposes.

Does KeptPDF upload my file?

No. Anonymize for AI runs in your browser, and the file is not uploaded to KeptPDF. The anonymized text you then paste into an AI tool does go to that tool. The names and numbers you swapped out stay on your computer.

Swap the names out before ChatGPT sees the file.

Open Anonymize for AI