ChatGPT, Copilot, Claude or Gemini: the plan matters more than the name

I’m not going to tell you which one to use. What matters is the plan you’re on, not the brand.

Each of the four comes two ways: a plan for one person, and a plan a business pays for. What happens to what you type can differ between them.

Which plan am I on?

Check the plan name in your account settings.

• ChatGPT. Personal: Free, Go, Plus and Pro. Business: Business.16
• Copilot. Personal: a personal Microsoft account. Business: a Microsoft 365 business account.79
• Claude. Personal: Free, Pro and Max. Business: Team.1419
• Gemini. Personal: a personal Google Account. Business: Google Workspace.2325

Signed up yourself, on your own card or for free? Assume it’s personal until the settings say otherwise.

The four questions to ask of your plan

• Step one. Which plan am I on: personal or business?
• Step two. Does it learn from what I type, and can I turn that off?
• Step three. Who can read what I type?
• Step four. How long is it kept, and where?

The answers are on each vendor’s privacy and help pages. Note what you find and the date, because the pages change.

Where we write “not stated”, the pages didn’t say. We’ll update this post when we find out.

What OpenAI’s pages say about ChatGPT

On a personal plan: Free, Go, Plus or Pro

• Does it learn from what I type? It can. OpenAI says “we may use your content to train our models” unless you turn it off.1
• How do I turn that off? Settings, then Data controls, then switch off “Improve the model for everyone”. It covers new chats only. Rate a reply thumbs up or down and that chat can still go into training, so skip the thumbs. A temporary chat skips your history and training, but may be kept up to 30 days for safety.2
• Who can read it? Not stated.
• How long, and where? Until you delete it, then up to 30 days, unless OpenAI must keep it for safety or legal reasons, or has removed your identity. Where: various countries, including the United States.3

On ChatGPT Business

• Does it learn from what I type? No, by default. OpenAI’s words: “We do not train our models on your data by default”.4
• Who can read it? Authorised OpenAI employees, for engineering support, abuse checks and legal compliance, and specialist contractors who review for abuse. Your own admins can view, export and delete staff chats, so tell your team first.4
• How long, and where? Your admin, the person who runs the account, decides how long. Deleted chats go within 30 days unless the law requires longer.4 Storage in Australia is only for some developer accounts and new Enterprise and Edu customers; Business isn’t named.5

What Microsoft’s pages say about Copilot

On a personal Microsoft account

Copilot at home has two rule books: one for the app updated from 18 August 2026, which the web version also follows, and one for the older app.910

• Does it learn from what I type? In the newer app, no. Microsoft says your prompts, its replies and your files “aren’t used to train foundation models”, the big AI underneath.9 The older app trains on your chats unless you opt out, and Australia isn’t on its list of countries left out.10 Copilot in Word, Excel and Outlook on Microsoft 365 Personal or Family isn’t used to train foundation models either.12
• How do I turn that off? The newer app has no training switch. Without paid Microsoft 365 you may see ads personalised from your chats and saved memories, which you can switch off in settings.13 In the older app, go to your profile, then Privacy (on the phone app, Account, then Privacy). Switch off “Training on conversation activity” and “Training on voice conversations”. That covers future chats only, and Microsoft can still use them for safety, security and advertising.11
• Who can read it? In the older app, some chats get human review, with no opt-out when Microsoft suspects you’ve broken its rules.10 For the newer app, not stated.
• How long, and where? In the older app, 18 months by default. The newer app’s pages give no time limit. You can delete chats in both.1113 Where: not stated.

On a Microsoft 365 work account

• Does it learn from what I type? Microsoft says your prompts, its replies and the work files and email Copilot reaches “aren’t used to train foundation LLMs”,7 LLMs being large language models, the AI underneath tools like Copilot. The free Copilot Chat on a work account makes the same promise.8
• Who can read it? Microsoft says Copilot has opted out of the human review it uses to watch for abuse.7
• How long, and where? With your business’s other Microsoft 365 content. Each person can delete their own history, and your admin sets how long it’s kept. Customers outside the European Union “may have their queries processed in the US, EU, or other regions”. Australia isn’t named.7
• What can it see? Your business’s email, calendar, documents, chats and meetings, limited to what each person can already open.7 As I read that, a file shared with everyone is one Copilot can find for everyone. Tidy the sharing first.

What Anthropic’s pages say about Claude

On a personal plan: Free, Pro or Max

• Does it learn from what I type? It can. Anthropic’s privacy policy says it may train on your chats “unless you opt out through your account settings”.15 Even with it off, chats flagged by its safety checks, or rated thumbs up or down, can be used.1416
• How do I turn that off? Settings, then Privacy, then switch off “Help Improve our AI models”. That stops future training, but anything already used stays used.16 An incognito chat isn’t used to improve Claude, even with the setting on.14
• Who can read it? If you allow training, Anthropic says it removes your identity first, and “Access is limited to a small number of personnel involved in model training”. Its help page says think twice before adding confidential business documents.18
• How long, and where? Until you delete it, then up to 30 days. If you allow training, chats may be kept, with your identity taken out, for up to 5 years. Chats flagged for breaking Anthropic’s rules may be kept for up to 2 years.17 Where: the United States and other countries.15

On Claude Team

• Does it learn from what I type? No, by default, except a chat you rate thumbs up or down. Anthropic keeps that for up to 5 years and may train on it. Your account owner can turn the thumbs off for everyone with the “Rate chats” setting.19
• Who can read it? The Primary Owner, the person who runs the account, can request an export of a member’s data, which may include their chats, so tell your team first.21 What Anthropic’s staff can read: not stated.
• How long, and where? Until someone deletes it, then up to 30 days. Chats flagged for breaking Anthropic’s rules are kept for up to 2 years.20 Anthropic says data is stored in the United States, and requests may be routed to the US, Europe, Asia and Australia.22

What Google’s pages say about Gemini

On a personal Google Account

• Does it learn from what I type? Yes, while a setting called Keep activity is on, which Google says is the default for anyone 18 or over.24 Google then uses your chats to improve its services, “including training generative AI models”.23
• How do I turn that off? At gemini.google.com, go to Settings and help, then Activity, then Turn off, or Turn off and delete activity.24 Chats are still kept up to 72 hours, to reply and for safety.2324 A temporary chat isn’t used for training.23
• Who can read it? People can. Google says human reviewers, some from outside companies, read some chats once disconnected from your account. It asks you not to enter anything confidential you wouldn’t want a reviewer to see. Reviewers can still check chats for safety with Keep activity off.23
• How long, and where? With Keep activity on, 18 months, unless you change it to 3 or 36 months, or never.24 Chats a human reviewer has seen are cut off from your account and kept for up to 3 years, even if you delete your activity.23 Where: not stated.

On Google Workspace

• Does it learn from what I type? No, without your permission. Google’s words: Workspace “does not use customer data for training models without customer’s prior permission or instruction”.25
• Who can read it? Google says human reviewers won’t read your chats or uploaded files without your permission. Other staff can’t see your prompts, even in a shared file.25
• How long, and where? Your admin decides whether chats are saved, and for 3, 18 or 36 months, with 18 the default. With saving off, chats are still kept for up to 72 hours.25 On Business Standard and higher, your admin can keep Gemini chats in the United States or Europe. Australia isn’t a choice.26
• What can it see? Only content each person can already open.25 As with Copilot at work, tidy the sharing first.

Where the vendors’ own pages don’t line up

OpenAI’s pages call one plan both ChatGPT Team and ChatGPT Business.146 Microsoft gives its work and home Copilot the same name, Microsoft Copilot.7

Anthropic’s help page says it trains on your chats if “You choose to allow us”.14 Its privacy policy says it may unless you opt out.15 Neither says whether the switch starts on or off for a new personal account.

If the makers can’t keep their own pages in step, check the plan and the setting, not the logo.

What Australian regulators say, whichever you pick

• The Office of the Australian Information Commissioner, the privacy regulator, recommends that “organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools”.27 That means AI tools anyone can sign up to online.
• The Australian Signals Directorate, the government’s cyber security agency, tells small businesses to “remove, anonymise or change personal details when using an AI application”.28 Anonymise means taking out the name and anything else that points to the person.
• The Tax Practitioners Board issued AI guidance on 22 July 2026 that binds registered tax and BAS agents, not every business.29 BAS is the Business Activity Statement, the form that reports your GST.

None of these call any of the four safe or unsafe. The first two say take the details out first. None of this is legal advice.

Want a second pair of eyes on your settings?

That’s what the free hour is for: we look at the plan you’re on, its settings, and what your staff put into it. Book a free hour.