You're offline, showing the last version we saved.
Build with AI · Using AI safely

Is ChatGPT safe? Privacy settings for ChatGPT, Claude, Copilot and Gemini

Same brand, different rules. Whether an AI tool learns from your chats depends on the plan you’re on and a few settings. Here’s where to find them in the four tools firms use most.

For beginners · Checked 11 November 2026 · Not advice

In plain English

On free and personal plans, most AI tools can use your chats to improve their models unless you switch it off. On business and enterprise plans, the providers say they don’t, by default.

So “is ChatGPT safe?” really means: which plan are you on, which settings have you changed, and what are you putting in?

Where the training setting lives

ChatGPT, personal plans
Settings, then Data controls, then switch off “Improve the model for everyone”. Voice mode has its own control. OpenAI says its Business, Enterprise and Edu plans don’t train on your data by default.
Claude, Free, Pro and Max
Settings, then Privacy: switch off the setting that lets Anthropic use your chats to improve Claude. Incognito chats aren’t used for training. Anthropic says its commercial plans, such as Team and Enterprise, aren’t trained on by default.
Microsoft Copilot, personal app
Privacy settings (on copilot.com, your profile name, then Privacy): switch off model training on text and on voice. Microsoft says work accounts signed in through your organisation aren’t used for training.
Gemini, personal Google account
Switch off Keep Activity in Gemini Apps activity. Chats are still kept for up to 72 hours and can be seen by human reviewers for safety. Google says Gemini in Google Workspace isn’t reviewed by people or used for training without permission.
Every tool: the feedback buttons
A thumbs up or down can send the whole conversation to the provider, even with training off. Give feedback in the chat instead.

Checked against each provider’s help pages on 11 November 2026. Menus move and apps get redesigned: if yours looks different, search the provider’s help for “model training”.

For beginners

Lock down your firm’s AI accounts in six steps

  1. Step 1: Find out who uses what, on which plan

    Ask everyone which AI tools they use for work and whether it’s a personal or business account. Same brand, different rules.

  2. Step 2: Switch training off on every personal account

    Each person does it on their own account, on web and phone. Voice settings can be separate.

  3. Step 3: Stop using the feedback buttons

    Thumbs up or down can share the chat. Say what was good or wrong in the chat instead.

  4. Step 4: Check memory, history and connectors

    See what the tool remembers about you and which apps it can reach. Turn off what you don’t use.

  5. Step 5: Move client work to a business plan, or redact first

    A business plan adds admin controls and a data processing agreement. Either way, strip client details first.

  6. Step 6: Check again every quarter

    Settings, plans and names change. Put a reminder in the diary.

Business plans

What a business plan adds, and what it doesn’t

Business and enterprise plans change the defaults, not your responsibility.

  • Training off by defaultThe providers say business data isn’t used to train their models unless you agree.
  • Admin controlsSomeone in the firm can manage accounts and settings for everyone, instead of trusting each person to do it.
  • A data processing agreementThe provider handles data on your instructions. You can show it to clients or a regulator.
  • It doesn’t make you compliantYour firm stays responsible for what goes in. Client details still need stripping unless your policy approves the tool for them.

Some business plans have a minimum number of seats. A sole practitioner can stay on a personal plan with training off and strict redaction.

Use these next

Please read: use these at your own risk

  • These prompts, guides and tools are general information, not professional, legal, privacy or IT advice. The Firm takes no responsibility for how they are built, changed or used, or for anything that results.
  • AI writes a slightly different tool each time. Whatever you build or download is yours to check, test and approve before you rely on it or use it with client data.
  • Check your AI provider’s current settings and terms for how your data is stored and whether it is used to train its models. Plans and settings change.
  • Bring your own judgement, and follow your firm’s AI policy and your professional and privacy obligations.

Questions firms ask

Is ChatGPT safe for accountants?

It can be used safely with the right plan and settings. Switch off “Improve the model for everyone” on personal accounts, avoid the feedback buttons, and keep identifiable client details out unless your firm has approved a business plan for them. OpenAI says its business plans don’t train on your data by default.

Does Claude train on my chats?

On Free, Pro and Max plans, only if the privacy setting that lets Anthropic use chats to improve Claude is on. Incognito chats aren’t used. Anthropic says commercial plans such as Team and Enterprise aren’t trained on by default.

Does Microsoft Copilot train on my data?

The personal Copilot app has privacy toggles for model training on text and on voice. Microsoft says Copilot used with a work account signed in through your organisation isn’t used to train its models.

Is Gemini safe for client data?

On a personal Google account, chats can be seen by human reviewers and, with Keep Activity on, used to improve Google’s products. Google says Gemini in Google Workspace isn’t reviewed by people or used for training without permission. Keep client details out of personal accounts either way.

Can I turn training off for my whole team at once?

On business and enterprise plans, an admin manages settings for everyone. On personal plans, each person has to change their own settings, which is one reason firms move to a business plan.