It depends on which ChatGPT plan you use and what you put into it. On ChatGPT Business and Enterprise, OpenAI says it doesn’t train on your data by default, encrypts it, and lets admins control retention. On personal plans, your chats can be used for training unless you switch that off. The plan is only half the answer: you still need clear rules on what staff never paste.
Last checked against OpenAI, ICO, Office of the Privacy Commissioner of Canada and FTC pages: 14 Oct 2026. This is general information, not legal advice.
Personal plans vs business plans
The biggest risk is usually staff using personal accounts for work. OpenAI’s own help pages and enterprise privacy page set out the differences:
- Training. For Free, Plus and Pro, OpenAI says “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. For Business, Enterprise, Edu and the API, OpenAI says it doesn’t use inputs or outputs to improve its models by default.
- The feedback catch. On personal plans, a thumbs up or down means “the entire conversation associated with that feedback may be used to train our models”, even after you opt out. On Business and Enterprise, OpenAI says feedback doesn’t permit training on that conversation.
- Retention. Personal chats stay in your account until you delete them. Deleted chats are scheduled for permanent deletion within 30 days, unless OpenAI must keep them for security or legal reasons. On Business and Enterprise, workspace admins control how long data is kept.
- Who can see chats. Teammates don’t see your Business chats unless you share them. Workspace admins, though, “can view, access, export, and delete end user conversations in the workspace”.
- Controls and paperwork. OpenAI’s pricing page lists SAML SSO, an admin console and SOC 2 Type 2 compliance for business plans, not personal ones. OpenAI will sign a Data Processing Addendum (DPA) for Business, Enterprise and the API.
For a small team, the practical step is simple: if staff use ChatGPT for work, move them into a Business workspace (available from 2 users) instead of personal accounts. Anyone who still uses a personal plan should keep the training opt-out switched off.
What not to paste: green, amber, red
A better plan doesn’t make every input safe. Sort what staff paste by how much harm a leak would do:
| Level | Examples | Rule |
|---|---|---|
| Green | Your published web pages, public product details, public reviews, blog outlines, emails with no names in them. | Fine on any plan. Check the answer before you use it. |
| Amber | Internal process notes, draft proposals, pricing plans, customer questions with the names removed, spreadsheets without personal data. | Business or Enterprise workspace only. Strip names, emails, phone numbers and account numbers first. |
| Red | Passwords and API keys, card or bank details, ID numbers (Social Security, National Insurance, SIN), health records, staff HR files, anything under an NDA, client data your contract doesn’t let you share. | Keep it out of ChatGPT unless a lawyer or data protection lead has approved a specific setup for it. |
This follows what regulators already ask of businesses. The FTC’s guide to protecting personal information says: “If you don’t have a legitimate business need for sensitive personally identifying information, don’t keep it. In fact, don’t even collect it.” The same logic applies to prompts: if the task works without the personal data, leave it out.
UK, Canada and US: the short version
UK (UK GDPR). The ICO’s AI and data protection guidance makes your business accountable, not the tool. It says that “in the vast majority of cases”, using AI with personal data is likely to be high risk and so needs a data protection impact assessment (DPIA), decided case by case. It also says that if you can’t tell whether a third-party AI tool would be compliant, you should, as good practice, “opt for a different solution”. The ICO notes this guidance is under review after the Data (Use and Access) Act.
Canada (PIPEDA). The Office of the Privacy Commissioner’s generative AI principles say businesses using these tools should, “where possible and reasonable, use anonymized or de-identified information within prompts”. Sensitive or confidential information should go into a prompt only “where authorised”.
US. The rules depend on your sector and state, but the FTC notes that laws such as the FTC Act and the Gramm-Leach-Bliley Act “may require you to provide reasonable security for sensitive information.” For patient data under HIPAA, OpenAI says it does not offer a Business Associate Agreement for ChatGPT Business. BAAs are available for the API and for sales-managed Enterprise or Edu accounts.
A five-rule AI policy for a small team
You don’t need a 20-page document. Five rules cover most of the risk:
- One approved workspace. Work happens in the company’s Business or Enterprise workspace, never on personal accounts.
- Pin the traffic-light table. Put it where people actually work, and say who to ask about anything amber or red.
- Lock sign-in. Turn on multi-factor authentication or SSO, both listed for Business, and remove people the day they leave.
- Control apps and GPTs. Admins choose which connected apps are switched on, and Business has controls for third-party GPTs. Only connect the tools staff need.
- Watch shared links. Business links open only inside your workspace, but you can’t switch link sharing off for everyone. That control is Enterprise-only, so tell staff to delete old links.
One more habit: treat every answer as a draft. Canada’s principles ask businesses to check that AI outputs are accurate enough for their purpose, especially before they’re published or used for decisions about people.
When to get help
Setting up a business workspace, a short AI policy and staff training is part of our AI adoption work for service businesses. If your question is the other ChatGPT question, whether it can find and recommend your business, read does ChatGPT use your Google Business Profile. To see where you show up on Google, Maps and AI answers today, run the free Search Gap check; we typically reply within 1–2 business days.