Merytt Talk through your business
AI

It depends on the plan, and on who picked it.

On free and personal ChatGPT plans, chats can help train OpenAI's models unless someone turns that off. Business plans work the other way round. Here is what the vendors and the Irish regulator actually say, and a one-page rule you can use this week.

By Aidan SmaleWritten Five minutes to read

On by defaultOn ChatGPT Free, Plus and Pro, sharing chats to train OpenAI’s models is switched on until someone turns it off, says OpenAI.
Off by defaultOn ChatGPT Business and Enterprise, OpenAI says it does not use your workspace content to train its models.
30 daysis how long OpenAI may keep a temporary chat for safety reasons, even though it is not used for training.
Controlleris what your firm could be when staff put personal data into an AI tool, says Ireland’s Data Protection Commission.

Is it safe for my team to put client information into ChatGPT?

It can be, but only on the right plan with the right settings. On free and personal ChatGPT plans, your chats can help train OpenAI’s models unless someone turns that off. On ChatGPT Business and Enterprise, OpenAI says it does not train on your data by default. Write a one-page rule and give staff an approved tool.

The real risk in most small firms is not the tool. It is that nobody decided which tool. So each person picks their own, on their own account, with their own settings.

If you have not chosen a tool for your team, your team has chosen several.

Does ChatGPT use what my staff type to train its models?

That depends on the plan they are signed into. OpenAI’s own help page is clear about the personal plans:

“On personal workspaces with ChatGPT Free, Plus, or Pro, data sharing is enabled by default.”

Anyone can turn it off. It is in Settings, then Data controls, then a switch called Improve the model for everyone. Once it is off, new chats are not used for training.

There is a catch. OpenAI says that if you give feedback, such as a thumbs up or thumbs down, the whole conversation may be used for training, even after you opt out.

The business plans work the other way round. OpenAI says it does not use content from ChatGPT Business, Enterprise, Edu or ChatGPT for Healthcare workspaces to train its models by default. ChatGPT Business is the plan that used to be called ChatGPT Team. OpenAI renamed it on 29 August 2025.

Is ChatGPT confidential if training is switched off?

Not in the way most people mean it. “Not used for training” and “nobody else can see it” are two different promises. The data is still sent to OpenAI and stored there for a time.

Here is what OpenAI says happens on ChatGPT Business:

  • Your admins can see it. Workspace admins can view, export and delete members’ conversations.
  • Some OpenAI staff can see it. Access is limited to authorised employees for support, abuse checks and legal reasons, and to contractors who review for abuse and misuse.
  • Deleting takes time. Deleted conversations are removed within 30 days, unless the law or safety needs mean keeping them longer.

Temporary chats on personal plans are not used for training either. But OpenAI says they may still be kept for up to 30 days for safety.

For a firm that needs a contract in place, OpenAI says it can sign a Data Processing Addendum for ChatGPT Business, ChatGPT Enterprise and its API.

Does Microsoft Copilot train on our company data?

Microsoft says no, for the work version. Its privacy page for Microsoft Copilot, the tool that until recently was called Microsoft 365 Copilot, puts it like this:

“Prompts, responses, and data accessed through Microsoft Graph aren’t used to train foundation LLMs, including those used by Microsoft Copilot.”

Two more points from the same page matter for an Irish firm:

  • It sees what each person can see. Copilot only shows a user the files and emails they already have permission to open. So messy sharing settings become a Copilot problem too.
  • EU data stays in the EU, mostly. Microsoft says EU traffic stays inside its EU Data Boundary. It also says models supplied by Anthropic inside Copilot are currently outside that boundary.

If your firm already runs on Microsoft 365, this may be the simplest approved option. Check with whoever manages your licences which Copilot your team actually has.

What does GDPR expect when staff use ChatGPT at work?

Ireland’s Data Protection Commission wrote about this in July 2024. Its advice to organisations is plain. If your staff put personal data into an AI tool, your firm could be a data controller, and a formal risk assessment should be considered.

Before using a tool, the DPC says you should understand:

  • what personal data it uses, and how;
  • where that data goes when a third party is involved;
  • whether the provider keeps it or reuses it;
  • how the tool lets you meet your GDPR duties.

It also asks a question worth putting to any tool. If someone asks you to show or delete the data you hold on them, can you do it inside the AI system?

GDPR is only part of it. A lot of client information is not about a person at all. It is pricing, drawings, plans and figures. Your contracts and NDAs may say where that can go. That part is our view, and worth checking against your own agreements.

What should a small business AI policy say?

It does not need to be long. One page that people actually read beats ten pages nobody opens. Here is a short rule you can copy, fill in and send round this week. It is our template, not a legal document.

  1. Our approved tool: [tool and plan, for example ChatGPT Business or Microsoft Copilot], signed in with your work account. Use this for anything to do with clients.
  2. Not for client work: personal accounts, free plans and any tool not on this page.
  3. Never paste: passwords or keys, health or money details about a person, ID numbers, staff HR matters, and anything a client shared with us in confidence or under an NDA.
  4. Fine to paste: our own drafts, public information, and text where names and details have been taken out.
  5. Before it goes out: a person reads every AI answer before a client sees it. The tool can be wrong.
  6. Feedback buttons: do not use thumbs up or down on chats that hold client details.
  7. Not sure? Ask [name] before you paste, not after.
  8. Reviewed: [date]. We check this page every six months, because the vendors change their terms.

Then do one more thing. Ask each person, with no blame, which AI tools they already use for work. The answer tells you where your client data has been going.

The rule is the easy part. The hard part is making the approved tool the easiest one to reach for. If the right tool is slower than a free tab in the browser, people will drift back. That is the kind of thing we help firms sort out, and we are always happy to talk it through.

Where this comes from.

Every rule and quote here is from the vendor or the regulator, opened on the day we checked. We do not write claims we cannot point at.

  1. Enterprise privacy at OpenAIOpenAI, updated 8 January 2026. No training on business data by default, who can see ChatGPT Business conversations, the 30-day deletion window and the Data Processing Addendum.
  2. Data controls in ChatGPTOpenAI Help Center. Training on by default for Free, Plus and Pro, how to turn it off, temporary chats, and the feedback catch.
  3. ChatGPT Business overviewOpenAI Help Center. Confirms ChatGPT Team was renamed ChatGPT Business on 29 August 2025.
  4. Data, Privacy, and Security for Microsoft CopilotMicrosoft Learn, updated October 2026. No training on prompts or Graph data, permissions, the EU Data Boundary and its Anthropic exception, and the rename from Microsoft 365 Copilot.
  5. AI, Large Language Models and Data ProtectionData Protection Commission (Ireland), 18 July 2024. What organisations should understand before staff use AI tools with personal data.

Last checked 28 October 2026. Vendors change their terms, so check the pages above before you decide. Nothing here is legal advice. If you handle sensitive or regulated client data, take advice from a data protection specialist.

Questions

Is it safe for employees to paste client data into ChatGPT?

It depends on the plan and the settings. On ChatGPT Free, Plus and Pro, chats can be used to train OpenAI’s models unless the user turns that off. On ChatGPT Business and Enterprise, OpenAI says it does not train on your data by default. Either way, decide what must never be pasted.

Is ChatGPT confidential?

Not fully. Switching off training is not the same as keeping data private. On ChatGPT Business, workspace admins can view conversations, and OpenAI says authorised staff and contractors may access them for support, abuse checks and legal reasons. Deleted conversations are removed within 30 days, with some exceptions.

Is ChatGPT Team the same as ChatGPT Business?

Yes. OpenAI renamed ChatGPT Team to ChatGPT Business on 29 August 2025. OpenAI says content in ChatGPT Business workspaces is not used to train its models by default.

Does using ChatGPT at work fall under GDPR?

It can. Ireland’s Data Protection Commission says that if staff put personal data into an AI tool, your organisation could be a data controller, and a formal risk assessment should be considered. You should know where the data goes and whether the provider keeps or reuses it.

What should a small business AI policy include?

Keep it to one page. Name the approved tool and plan, say that personal accounts are not for client work, list what must never be pasted, require a person to check every answer before a client sees it, name who to ask, and set a review date.

Pick the tool before your team does.

Thirty minutes with a founder. Tell us which tools your team already uses, and we will tell you straight what we would change.

Or take the six-question diagnostic