Skip to content
Zrico
Menu

For every firm

Is ChatGPT Safe for Client Data?

How OpenAI handles your data in personal ChatGPT plans versus ChatGPT Business, Enterprise and the API, and what that means for client confidentiality.

Updated 7 min read

“Is ChatGPT safe?” has no single answer, because ChatGPT is several products with different data terms. What matters for a law or accounting firm is which version you use, how it is set up, and what your professional rules require. This guide sets out what OpenAI’s own pages say, as of October 2026.

Two families of products

OpenAI separates its products into two groups.

Services for individuals. These are the personal ChatGPT plans. OpenAI’s help center names Free, Plus and Pro.

Business products. OpenAI’s enterprise privacy page covers ChatGPT Business, ChatGPT Enterprise, ChatGPT for Healthcare, ChatGPT Edu, ChatGPT for Teachers and the API Platform. An organization runs these as a managed workspace or developer account.

The data rules differ between the two groups. The rest of this guide takes them in turn.

Personal plans: training is on unless you turn it off

OpenAI’s help center says that when you use its services for individuals, it may use your content to train its models. It also says it takes steps to reduce the amount of personal information in training datasets.

You can opt out. In ChatGPT, open Settings, then Data controls, and turn off “Improve the model for everyone”. You can also choose “Do not train on my content” in OpenAI’s Privacy Portal. Either one is enough. After you opt out, OpenAI says it will not use your new conversations to train its models.

There is one exception to know about. On personal plans, if you give feedback on a response, such as a thumbs up or thumbs down, OpenAI says the entire conversation tied to that feedback may be used for training. That applies even if you have opted out.

Temporary chat

Personal plans also offer temporary chat. According to OpenAI, a temporary chat:

  • does not appear in your chat history
  • does not create or update memories
  • is not used to improve OpenAI’s models while it stays temporary
  • may be kept by OpenAI for up to 30 days for safety purposes

If you save a temporary chat, it becomes a regular chat and follows your normal settings.

Deleting chats

When you delete a saved chat, OpenAI says it disappears from your account right away and is scheduled for permanent deletion from its systems within 30 days. The exceptions are chats that were already de-identified and disassociated from your account, and cases where OpenAI must keep data longer for security or legal reasons. Regular and archived chats stay saved until you delete them.

Business products: no training by default

OpenAI’s enterprise privacy page says that by default it does not use business data to train its models. The exception is if you explicitly opt in to share data, for example through opt-in feedback mechanisms. For the API, the default covers data sent after March 1, 2023.

On business plans, feedback works differently. OpenAI’s help center says thumbs-up or thumbs-down feedback in ChatGPT Business, Enterprise or Edu does not permit training on the conversation. Workspace admins can also turn response feedback off.

OpenAI’s enterprise privacy page also says:

  • As between you and OpenAI, you keep your rights in your inputs and own the output you rightfully receive, to the extent the law allows.
  • Data is encrypted at rest (AES-256) and in transit (TLS 1.2 or higher).
  • Business data may be run through automated safety classifiers. The resulting classifications are metadata and do not contain the business data itself.
  • OpenAI can sign a Data Processing Addendum for ChatGPT Business, ChatGPT Enterprise and the API.
  • ChatGPT Business, ChatGPT Enterprise and the API Platform have each completed a SOC 2 Type 2 audit.

ChatGPT Business

OpenAI describes ChatGPT Business as built for small and growing businesses. OpenAI says workspace admins can control how long data is kept. Deleted or unsaved conversations are removed within 30 days, unless the law requires longer retention or it is reasonably necessary to protect OpenAI’s services or a third party from harm.

On access, OpenAI says its people can see stored conversations only for engineering support, investigating possible abuse, and legal compliance. Specialized third-party contractors bound by confidentiality obligations may also review content, solely for abuse and misuse.

Each member has a private chat history. Shared links work only inside the workspace. OpenAI’s help center notes that a Business workspace cannot turn off shared links for everyone; that control is in Enterprise.

ChatGPT Enterprise

ChatGPT Enterprise adds more control. OpenAI says workspace admins control retention, and deleted conversations are removed within 30 days unless OpenAI is legally required to keep them. Admins can reach an audit log of conversations through OpenAI’s Compliance API. OpenAI says its employees access Enterprise conversations only to resolve incidents, to recover conversations with your explicit permission, or where the law requires.

The API

The API is for building your own applications. OpenAI’s platform documentation says API data is not used for training unless you opt in. By default, abuse monitoring logs, which can include prompts and responses, are kept for up to 30 days, unless the law requires longer or it is reasonably necessary to protect OpenAI’s services or a third party from harm. Eligible customers can apply for Zero Data Retention or Modified Abuse Monitoring, which exclude customer content from those logs.

What this means for client confidentiality

No training is an important protection, but it does not settle the question by itself. Your professional rules ask broader questions.

For lawyers. ABA Formal Opinion 512 says that before entering information relating to a representation into a generative AI tool, you must assess the risk that people outside the firm could access it. For “self-learning” tools, it says you need the client’s informed consent first, and that engagement-letter boilerplate is not enough. It also says you should read the tool’s terms of use and privacy policy, or rely on someone who has.

A personal plan with training turned on fits the risk the opinion describes. A personal plan with training turned off is better, but it is still an individual’s account, not the firm’s. The firm has no admin control, no business contract and no say over what happens when the employee leaves.

For accountants. The AICPA Code says a member in public practice should not disclose confidential client information without the client’s specific consent. Before sharing it with a third-party service provider, the member should either have a confidentiality agreement with the provider, with reasonable assurance about its safeguards, or get the client’s specific consent. A business plan with a signed agreement can support the first route. A personal account generally cannot. Tax return information has additional federal rules, covered in a separate guide.

For everyone. Even on a business plan, you should know:

  • how long data is kept, and who sets that
  • who at the provider can see it, and why
  • whether your plan lets you enforce the controls you need, such as turning off shared links
  • what your contract and Data Processing Addendum actually say

A sensible approach

  1. Do not use personal ChatGPT accounts for client work. Say so in your AI policy.
  2. If you want to use ChatGPT, use a business plan the firm controls, and sign the Data Processing Addendum.
  3. Set retention in the admin console to match your records policy.
  4. Decide which client information may be entered, and when you need client consent.
  5. Re-read OpenAI’s enterprise privacy page and help articles at least twice a year. They change, and the dates on them show when.

How Zrico fits

Zrico gives each firm its own private workspace, with its own application and databases. It runs Claude by default, processed only in the US. The firm owner can also turn on OpenAI’s GPT models and Google’s Gemini, whose processing is not limited to the US. No AI provider trains on prompts, files or answers. Zrico has no SOC 2, ISO or HIPAA certification, does not make a firm compliant, and its output must be reviewed by a professional.

Sources

General information, not legal advice. Check the current rules in your jurisdiction.

Give your firm a private place to use AI.

$39 per seat per month after a 30-day free trial, 5-seat minimum. Your workspace is ready minutes after you sign up.

Your card is charged when the trial ends. Cancel any time before then and you pay nothing.

Start 30-day free trial