Short answer: Not for client work. A personal ChatGPT, Claude or Gemini account belongs to the staff member, not the firm. On personal ChatGPT plans, data sharing for training is on by default, the firm can't see or delete what went in, and access doesn't end when the person leaves. Business plans from the same companies are different: they say they don't train on your data by default, and your admins control users and retention. If staff need AI for client work, give them a firm account on a plan whose terms you have read.
What is the difference between a personal and a business plan?
| Personal plan (e.g. ChatGPT Free, Plus or Pro) | Business plan (e.g. ChatGPT Business, Claude Team) | |
|---|---|---|
| Who owns the account | The staff member | The firm |
| Training on your content | ChatGPT: "data sharing is enabled by default" until turned off | ChatGPT Business: "By default, we do not use your business data for training our models." Claude Team: "No model training on your content by default" |
| Who controls retention | The staff member | Your admins, within the vendor's limits |
| What happens when they leave | They keep the account and its history | You remove their access |
| Single sign-on | No | ChatGPT Business: "Secure workspace with SAML, SSO, and MFA". Claude Team: "Single sign-on (SSO)" |
Sources: OpenAI data controls FAQ, OpenAI enterprise privacy, OpenAI pricing, Claude pricing, all read November 16, 2026. For more on whether ChatGPT trains on your data, see Is ChatGPT private? on Zero Data Retention AI.
How common is it?
Very. Cyberhaven, a data-security vendor, reports that "32.3 percent of ChatGPT usage happens through personal accounts" and 58.2 percent of Claude usage. It adds that "Personal accounts bypass SSO enforcement, centralized logging, enterprise retention policies, and controls related to data usage or model training" (Cyberhaven, February 2026). Those figures come from the businesses Cyberhaven studied, not from small firms in particular.
"But I turned training off"
Some careful staff do switch off "Improve the model for everyone". That helps, but it leaves the other problems:
- The chats are still kept in their account until they delete them.
- The firm can't check. You have no way to confirm the setting, or what went in.
- It doesn't end. When they leave, the client data in their chat history leaves with them.
- It may not meet your rules. If your sector expects you to choose and oversee the vendors that handle client data, a personal account is a vendor the firm never chose. See Shadow AI risks by sector.
What about paying for staff's personal Plus or Pro accounts?
Paying for a personal plan doesn't make it a business plan. On OpenAI's own pages, ChatGPT Plus and Pro are personal plans with data sharing on by default, and "Business and Enterprise are for businesses" (OpenAI pricing). If you are going to pay, pay for a plan your firm controls.
What to put in your policy
Two sentences cover it:
- "Personal AI accounts, including paid ones, must not be used for firm or client work."
- "Use only the firm's approved tool, signed in with your firm account."
The full wording is in our AI acceptable use policy template. For choosing the firm tool, see What makes an AI tool safe to approve for your staff?