ChatGPT can be safe for business data, but it depends on which version your team uses. On OpenAI's business plans (ChatGPT Business, Enterprise and the API), OpenAI says it doesn't train its models on your data by default, and admins control how long it's kept. On the individual plans (Free, Plus and Pro), OpenAI may use conversations to train its models unless each person turns that off. The bigger risk for most small businesses isn't OpenAI. It's staff pasting customer data, contracts or passwords into personal accounts nobody controls. The fix is simple: use a business plan, set a short AI policy, and keep certain information out entirely.
What OpenAI does with your data, by plan
These are OpenAI's own published policies as of this writing. Policies change, so check them before you decide.
| Free, Plus, Pro (individual) | ChatGPT Business, Enterprise | API | |
|---|---|---|---|
| Used to train models? | May be, unless the user opts out | Not by default | Not by default |
| Who controls it? | Each individual user | Your workspace admins | Your organization |
| Retention | Saved chats stay until deleted | Admins control retention; deleted chats removed within 30 days | Inputs and outputs kept up to 30 days, then removed |
| Admin visibility | None for your company | Admins can view, export and delete workspace conversations | Your systems |
Sources: OpenAI's enterprise privacy page and its help article on how your data is used to improve model performance.
Individual plans (Free, Plus, Pro)
OpenAI says that on its services for individuals, "we may use your content to train our models." You can opt out by turning off Improve the model for everyone under Settings → Data controls, and after that OpenAI won't use your new conversations for training. There's an important catch: if you give feedback on a response, such as a thumbs up or thumbs down, OpenAI says it may use the entire conversation associated with that feedback to improve its models, even if you've opted out (OpenAI).
Temporary chats aren't used to train models and don't appear in your history, but OpenAI may keep a copy for up to 30 days for safety purposes (OpenAI).
The bigger problem with individual plans in a business isn't training. It's that your company has no control. You can't see what was shared, you can't enforce settings, and when an employee leaves, the conversations, including anything they pasted in, leave with their personal account.
Business plans (ChatGPT Business and Enterprise)
On OpenAI's business plans, the company says: "By default, we do not use your business data for training our models." Workspace admins control how long data is retained, and deleted conversations are removed from OpenAI's systems within 30 days unless longer retention is legally required. Admins can view, export and delete conversations in the workspace, and OpenAI states that ChatGPT Business completed a SOC 2 Type 2 audit (OpenAI).
For most small businesses, moving the team onto a business plan is the single biggest improvement you can make. It turns a scattered set of personal accounts into something you manage.
The API
If you build ChatGPT's models into your own software or automations through the API, OpenAI says it doesn't train on that data by default and may keep inputs and outputs for up to 30 days to provide the service and identify abuse. OpenAI also says it can sign business associate agreements (BAAs) for customers who need HIPAA compliance, on request (OpenAI). If you handle patient data, see our guide to AI for dental and medical practices.
The real risks for a small business
- Personal accounts used for work. The most common problem by far. Staff sign up for free accounts, paste in customer emails, spreadsheets or contracts, and nobody knows it's happening.
- Sensitive data that shouldn't go anywhere. Passwords, API keys, bank and card numbers, and Social Security numbers should never be pasted into any AI tool. Leaked keys are one of the fastest ways to get breached, as the Railway incident showed.
- Shared links. ChatGPT lets people share a conversation by link. Anyone with the link can see it, so a shared chat containing client details is a leak.
- Connected apps and custom GPTs. Connecting ChatGPT to email, drives or other apps, or using a custom GPT that calls outside services, means data can flow to more places. OpenAI notes that data sent to a third party through a GPT's actions is governed by that third party's privacy policy (OpenAI).
- Regulated data. Health, financial and some legal or government data come with their own rules, and a consumer account almost never meets them.
- Trusting the answer. Not a privacy risk, but a business one: AI can be confidently wrong. Anything that goes to a customer, into a contract or into your books needs a person to check it.
A simple AI policy you can use this week
You don't need a 20-page document. One page covering these points is enough for most small businesses:
- Approved tools. List the AI tools staff may use for work, on company accounts only. No personal accounts for company work.
- Never share: passwords, API keys, card or bank numbers, Social Security numbers, patient health information (unless you have a BAA and approved tool), or anything covered by a client confidentiality agreement.
- Okay with care: customer names and business details on an approved business plan, using only what the task needs.
- Always fine: public information, general questions, drafting from scratch, and anonymized examples.
- Review before it goes out. A person checks anything AI drafted before it reaches a customer, a contract or your accounting.
- No shared links for conversations that contain anything about customers or the business.
- Ask first before connecting AI tools to email, drives, the CRM or other systems.
- Who to ask. Name one person who answers questions and approves new tools.
Then check the settings: on a business plan, set retention to what you need and confirm which apps and connectors are allowed. On any individual accounts that remain, have users turn off model training.
The same questions apply to every AI tool
Claude, Microsoft Copilot, Google Gemini and the AI features built into tools you already use all have their own policies. Before your team uses any of them for work, ask:
- Is our data used to train models, and is that off by default on our plan?
- How long is data kept, and who can delete it?
- Can we manage accounts centrally and see what's being shared?
- Will the vendor sign the agreements our industry needs, like a BAA?
- What connects to what, and where does the data go?
Our AI readiness checklist includes these as part of a broader look at your data, tools and security, and our guide to what AI automation costs covers what business AI seats cost.
Get it set up right
Our AI Readiness Audit includes exactly this: which tools your team is already using, where AI is safe to use with your data, where a person must stay in the loop, and the policies you need, alongside a ranked, costed plan of what to automate. The price is fixed and agreed before we start.
Book a free call and tell us how your team uses AI today. We'll tell you what we'd tighten up first.