
Does ChatGPT use your data for training? On the personal plans (Free, Go, Plus and Pro), yes by default: OpenAI uses your conversations to train its models unless you turn off "Improve the model for everyone" under Settings, then Data controls. On ChatGPT Business (formerly ChatGPT Team) and ChatGPT Enterprise, no: workspace data is not used for training by default, and you own your inputs and outputs. Claude follows the same split, with one difference: its personal plans (Free, Pro and Max) use chats for training only when the model-improvement setting in your account allows it, while Claude Team, Enterprise and API data is not used for training by default. The rest of this FAQ covers retention, who can see what, Microsoft 365 Copilot, and what staff may enter under an approved policy, as of October 2026.
If your staff handle patient information, the question changes; see HIPAA rules for ChatGPT and other AI tools. For the workspace settings themselves, see the ChatGPT Business security settings and data privacy checklist. The quickest fix for most of these privacy questions is a written AI acceptable use policy, backed by controls that keep client data out of unapproved tools.
Does ChatGPT use your data for training on each plan?
The answer depends on the plan, not on how careful the user is. OpenAI's enterprise privacy page states that data from ChatGPT Business, ChatGPT Enterprise and the API is not used for training by default unless you explicitly opt in, and its pricing pages mark the personal plans as "Opt-out available." Anthropic's privacy center says the same for Claude's commercial products. Here is the picture as of October 2026:
| Plan | Trains on your data by default? | Retention and deletion | Who controls it |
|---|---|---|---|
| ChatGPT Free, Go, Plus, Pro (personal) | Yes, unless you turn off "Improve the model for everyone" in Settings, Data controls. Thumbs up or down feedback can still send the whole conversation for training. | Deleted chats are scheduled for permanent deletion within 30 days. Temporary chats stay out of history and are not used to improve models, but OpenAI may keep a copy for up to 30 days for safety. | The individual user |
| ChatGPT Business (formerly Team) | No. OpenAI states it will not train on workspace data, including memories, data reached through connected apps, GPT conversations and projects. | Workspace admins can control how long data is retained; deleted or unsaved conversations are removed within 30 days. When a member is removed, their chats, files and canvas documents are retained indefinitely and restored if they are re-added. | Workspace owners and admins; no BAA is offered on Business |
| ChatGPT Enterprise | No, by default. | Admins set custom retention policies; deleted conversations are removed within 30 days; admins can reach an audit log of conversations through the Compliance Platform. | Enterprise admins, with SCIM, IP allowlisting and data residency available |
| OpenAI API | No, unless you opt in. | Abuse-monitoring logs are kept for up to 30 days and stored API responses default to 30 days; zero data retention is available by approval. | Your developers and the API organization owner |
| Claude Free, Pro, Max (personal) | Only when you allow Anthropic to use your chats and coding sessions to improve Claude, when a chat is flagged for safety review, or when you otherwise opt in. Incognito chats are never used. | With model improvement on, chats may be kept de-identified for up to 5 years; deleted conversations leave back-end storage within 30 days. | The individual user |
| Claude Team and Enterprise | No, by default. Training use is opt-in through feedback or explicit permission. | Deleted chats leave history immediately and back-end storage within 30 days; Enterprise adds custom data retention controls. | Organization owners; Enterprise can accept a BAA |
| Claude API | No, by default. | Inputs and outputs are deleted within 30 days unless a feature you control stores them; zero data retention is available through sales. | Your developers and the organization owner |
| Microsoft 365 Copilot | No. Prompts, responses and data accessed through Microsoft Graph are not used to train foundation models. | Your data stays inside your tenant's protections and existing permissions. | Your Microsoft 365 admins |
Sources for the table: OpenAI's data controls article and ChatGPT Business overview, Anthropic's consumer and commercial privacy articles, and Microsoft's Copilot privacy documentation, all opened in October 2026.
How do you stop ChatGPT from training on your data?
- On a personal account, turn the setting off. Open Settings, then Data controls, and switch off "Improve the model for everyone." Do it on every device and account your staff use. Avoid the thumbs up and thumbs down buttons on anything sensitive, because feedback can send the whole conversation for training even after you opt out.
- Use Temporary chat for one-off sensitive questions. A temporary chat stays out of your history, does not create memories and is not used to improve models while it is temporary, though OpenAI may keep a copy for up to 30 days for safety.
- Move company work into a business workspace. ChatGPT Business starts at two seats, does not train on workspace data by default, and gives owners control over memory, apps and retention. Our guide to moving personal ChatGPT use into a company workspace covers the merge, and ChatGPT Business vs Plus and Pro explains what else changes.
- Write the rule down. A one-page policy that names the approved account type for each kind of data does more than any setting, because settings live on devices you do not control.
Does Claude use your data for training?
Not on business plans by default. Anthropic's privacy center, updated August 18, 2026, says it will not use inputs or outputs from its commercial products (Claude Team, Claude Enterprise, the API, and Claude Code on those accounts) for training unless you opt in through feedback, bug reports or explicit permission. On the personal plans (Free, Pro and Max), your chats and coding sessions are used for training when you choose to allow Anthropic to use them to improve Claude, when a conversation is flagged for safety review, or when you otherwise opt in; incognito chats are never used and raw content from connectors is excluded. Anthropic's article does not state which way that setting is set when an account is created, so check it in your own account rather than assuming.
Retention follows the same pattern. On commercial plans, deleted chats leave history immediately and back-end storage within 30 days, and feedback submitted with a thumbs up or down is stored for up to 5 years, de-linked from your account; owners can turn that rating feature off under Organization settings, Data and Privacy. On personal plans with model improvement on, chats may be kept in de-identified form for up to 5 years. Claude Enterprise adds custom data retention controls and a HIPAA-ready configuration with a BAA; Claude Team does not. For a small business weighing the two, our Claude for small business guide covers setup and security.
What are the privacy and security concerns with ChatGPT at work?
Owners tend to worry about the model "learning" their secrets. Once the plan is right, that is the smallest of five concerns:
- Personal accounts. Staff using Free or Plus accounts for work are on the plan where training is on by default, and the company has no visibility, no retention control and no way to remove the data when someone leaves.
- Connected apps. In ChatGPT Business, apps that connect Google Drive, SharePoint, Slack and similar services are enabled by default, and admins decide workspace-wide availability. Each app reaches whatever the signed-in user can already see, so overshared files become answerable questions.
- Who can see conversations. OpenAI's enterprise privacy page says Business workspace admins can view, access, export and delete end-user conversations, while its Business help center says admins and owners cannot see private member chats by default and that owners export a member's data from the Members page. Tell staff plainly which applies in your workspace, and treat work chats as company records either way.
- Retention and departures. Business keeps a removed member's chats, files and canvas documents indefinitely unless they are deleted. That is useful for continuity and a liability if nobody owns the cleanup.
- Accuracy and disclosure. Confident wrong answers, pasted client details that end up in a draft sent to the wrong client, and work product nobody told the client was AI-assisted are ethical concerns rather than technical ones, and a usage policy is where they get settled.
The practical response is the same list every time: business accounts, a short policy, training on what to paste and what not to, and data loss prevention watching the exits. Our shadow AI guide covers the personal-account problem in depth.
Is ChatGPT Business private enough for client data?
For most small businesses, yes, with the settings reviewed and a policy in place. ChatGPT Business (the plan many people still search for as ChatGPT Team) does not train on workspace data by default, has completed a SOC 2 Type 2 audit, supports a Data Processing Addendum for GDPR and similar laws, includes SAML and OIDC single sign-on with domain verification, and lets owners control memory and app availability. What it does not include, as of October 2026: a Business Associate Agreement for HIPAA ("We do not offer a BAA for ChatGPT Business," per OpenAI's help center), SCIM provisioning, role-based access controls, the Compliance Platform for log export, enterprise key management, IP allowlisting, or the data residency that OpenAI's pricing table lists for Enterprise only (its help center says Business data residency is rolling out gradually and is chosen at checkout where available). If your clients send security questionnaires that ask for those, compare ChatGPT Business vs Enterprise before you buy seats; an upgrade that keeps your workspace data goes through OpenAI sales.
What may staff enter into an AI tool under an approved policy?
A policy that only says "be careful" fails on the first busy afternoon. Tie each class of data to an account type. This is the structure we use; adjust the classes to your own business and have counsel review the result.
| Data class | Personal ChatGPT or Claude account | Approved business workspace (ChatGPT Business, Claude Team or Enterprise, Copilot) |
|---|---|---|
| Public information (your website, published prices, general research) | Allowed | Allowed |
| Internal, non-sensitive work (drafts, procedures, meeting notes without client names) | Not allowed | Allowed |
| Client-confidential material (contracts, financials, correspondence) | Not allowed | Allowed once the training default, retention setting and contract terms have been checked and recorded for that tool |
| Regulated data (patient information, card numbers, data under a confidentiality agreement that names AI) | Never | Only on a plan covered by the required agreement, such as a BAA, and only where the policy names it; ChatGPT Business is not eligible for a BAA |
| Passwords, keys, tokens, unreleased source code | Never | Never, except source code inside a tool your developers were approved to use for it |
Our AI usage policy guide has a template with these tiers built in.
Frequently asked questions
Is my business data safe when we use AI tools?
It depends entirely on which tool and which account. A business-tier AI service with the right agreement keeps your data private and out of model training. A free consumer account makes fewer promises. The safety comes from the version you use and the settings you apply, not from the word "AI" itself.
Does ChatGPT Plus use your data for training?
Yes, by default. ChatGPT Plus is a personal plan, and personal plans use conversations for training unless the user turns off "Improve the model for everyone" in Settings, Data controls. Paying for Plus does not change that default; moving the work into ChatGPT Business or Enterprise does. If an employee uses Plus for work, assume the training setting is on until you have seen it switched off.
Does the free version of ChatGPT use what we type?
Yes, unless the setting is turned off. By default, personal ChatGPT accounts, including Free, may use your conversations to improve OpenAI's models, and the same is true of most consumer chatbots. That is the core reason a free personal account is a poor place for anything confidential. Assume a consumer tool keeps what you paste unless you have confirmed otherwise.
Does deleting a ChatGPT chat remove it from training?
Deleting a chat schedules it for permanent deletion from OpenAI's systems within 30 days, and it cannot be restored. Deleting does not change the training setting, so switch "Improve the model for everyone" off first; from then on, new conversations on that account are excluded. Business and Enterprise workspaces are excluded by default, so the question only arises on personal plans.
Does Microsoft 365 Copilot train its AI on our company data?
No. Microsoft states plainly that "prompts, responses, and data accessed through Microsoft Graph aren't used to train foundation LLMs," according to its Copilot privacy documentation. Your data stays inside your tenant's protections and existing permissions. Our Copilot FAQ covers what it can and cannot do in more depth.
Can our admins read employees' ChatGPT Business chats?
OpenAI's two official pages describe it differently, so quote both in your policy. The enterprise privacy page says workspace admins can view, access, export and delete end-user conversations. The Business help center says each member has their own chat history, that admins and owners cannot see all private member chats by default, that usage analytics do not expose transcripts, and that owners can export a member's data from the Members page when that option is available. Either way, staff should treat work chats as company records, and shared links in Business can only be opened by members of the same workspace.
How long does ChatGPT keep our data?
On Business and Enterprise, workspace admins control the retention period, and deleted or unsaved conversations are removed from OpenAI's systems within 30 days unless the law requires longer. On personal plans, deleted chats are scheduled for permanent deletion within 30 days, and temporary chats may be kept for up to 30 days for safety. Knowledge files attached to a custom GPT are kept until the GPT is deleted.
What is the real risk with staff using AI at work?
The risk is quiet data leakage, not a dramatic hack. LayerX's 2025 enterprise report found that about 18% of employees paste data into generative AI tools, and more than half of those pastes contained corporate information, with 77% of that AI use happening through personal accounts outside company controls, as eSecurity Planet reported. That is your data leaving unseen.
What data should never go into a public AI tool?
Keep client records, contracts, financials, passwords and keys, source code, employee information, and anything covered by a confidentiality agreement out of consumer AI accounts. A useful test: if you would not post it in a public forum, do not paste it into a free chatbot. For work with that kind of data, use an approved business-tier tool instead.
What is the difference between a consumer AI account and a business one?
A business or enterprise tier comes with a contract that governs how your data is handled, a no-training default, admin controls over members, memory and apps, retention you can set, and usually compliance coverage such as a SOC 2 report or a DPA. A free consumer account is governed by general terms written for the public, with training on by default. Same underlying models in many cases, very different promises about your data.
Can we let staff use AI safely?
Yes, with a little structure. Pick approved tools, use business tiers for anything touching company data, write a short usage policy so people know the rules, and turn on data loss prevention to catch sensitive information leaving. Our AI usage policy guide and data loss prevention FAQ give you the starting pieces.
Who owns the content an AI tool generates for us?
On ChatGPT Business, ChatGPT Enterprise and the OpenAI API, OpenAI's enterprise privacy page says you own your inputs and outputs where allowed by law. Other providers put the same point in their own agreements, so read the one you sign. The bigger practical concern is accuracy: AI can produce confident, wrong answers, so treat its output as a first draft that a person reviews, not a finished product to publish as-is.
By Jamie Baum for The NetSys Group, which has delivered managed IT, cybersecurity and cloud services from Brooklyn since 1998. Our engineers hold degrees in electrical and computer engineering. We work on-site across New York City, Long Island, Westchester, the Hudson Valley, North Jersey and Fairfield County, CT, and remote-first everywhere else.
What NetSys does
NetSys sets up ChatGPT Business, Claude Team and Claude Enterprise workspaces for small businesses with the training, memory, retention and app settings checked and recorded before anyone pastes a client file, writes the usage policy with you, and puts the controls behind it: an approved-tool list, browser and data loss prevention rules, and staff training. See ChatGPT Business setup and AI governance consulting. Request an AI data review and we will go through which accounts your staff actually use and what to change first.
Related reading
Also on this topic: AI Contract Review for Small Business: What to Trust · AI Policy Template: Write an AI Usage Policy Your Team Will Follow
Discuss chatgpt business setup for your business.
Tell us about your current systems, the result you need and your timeline. We will discuss the work, responsibilities and pricing before you decide on an engagement.



