Search Claude vs ChatGPT and most comparisons rank the two bots on writing and code. For an Australian business, the first question is which plan keeps your customers’ details out of model training. On both companies’ personal plans, your chats can help train future models while the training setting is on, and on ChatGPT it starts on. On their business plans, neither company trains on your data by default. Choose the plan and the setting first, then choose the model on features.
Full disclosure: we run our own agentic workflow on Claude Code, so we aren’t neutral. Everything below comes from the two companies’ own policy pages and Australia’s privacy regulator, not our taste.
Key takeaways
- On personal plans (ChatGPT Free, Plus and Pro; Claude Free, Pro and Max), your chats can be used for training while the setting is on. ChatGPT starts with it on.
- On business plans (ChatGPT Business and Enterprise; Claude Team and Enterprise) and on both APIs, neither company trains on your data by default.
- The thumbs up and thumbs down buttons can send a whole conversation to the vendor, even after you’ve opted out.
- The OAIC’s best-practice advice is to keep personal information out of publicly available AI tools altogether.
Claude vs ChatGPT on privacy, plan by plan
Checked against each company’s own pages on 11 November 2026. Policies change, so recheck before you set up a team.
ChatGPT | Claude | |
|---|---|---|
Personal plans | Free, Plus, Pro | Free, Pro, Max |
Can personal chats train models? | Yes. Data sharing is on by default | Yes, if you allow it (you choose at sign-up), or if a chat is flagged for safety review |
Where the switch lives | Settings, Data controls, Improve the model for everyone | Settings, Privacy, the model improvement toggle |
A chat that never trains | Temporary chat (a copy may be kept up to 30 days for safety) | Incognito chat |
Thumbs up or down | Can send the whole conversation into training, even after you opt out | Stores the whole conversation for up to five years; may be used for training |
Business plans | Business, Enterprise, Edu | Claude for Work (Team, Enterprise) |
Do business plans train? | Not by default, API included | Not by default, API included |
Checked against OpenAI’s and Anthropic’s help and privacy pages on 11 November 2026.
Sources: OpenAI’s help page, and Anthropic’s consumer and commercial privacy articles.
One row catches people out. Claude’s consumer rules changed in 2025. In an update published 28 August 2025, Anthropic started asking Free, Pro and Max users to choose whether their chats could train its models, and set retention at five years for anyone who says yes. If you set up Claude before then, or clicked through that pop-up in a hurry, check what you chose.
Switching off later doesn’t pull back what’s already gone in. Anthropic says your data stays in training runs already started and models already trained. Set the toggle on day one.
Why does the feedback button matter so much?
On ChatGPT’s individual plans, OpenAI says a thumbs up or thumbs down means the entire conversation tied to that feedback may be used to train its models, even if you’ve opted out. On Claude, a rating stores the whole conversation for up to five years, and Anthropic may use it to train its models. That applies on consumer and commercial plans alike.
Rating a helpful reply in a chat that holds a client’s details sends that chat to the vendor.
Business plans close the gap. On ChatGPT Business and Enterprise, OpenAI says feedback doesn’t permit training on the conversation, and workspace admins can turn response feedback off. On Claude Team and Enterprise, owners can switch off Rate chats in the organisation’s Data and Privacy settings.
What does the OAIC expect when your business uses AI chat?
The Office of the Australian Information Commissioner (OAIC) is direct about it. Its guidance on privacy and commercially available AI products says “the OAIC recommends that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools”.
It also says the Privacy Act applies to all uses of AI involving personal information. Its main asks:
- Do due diligence before you adopt a tool, including who can access what your staff type in.
- Read the terms and settings, especially whether the vendor can use your inputs to train or improve its models.
- Use personal information only for the purpose you collected it for. That’s Australian Privacy Principle 6. A second use needs consent or another exception, such as the customer reasonably expecting it.
- Update your privacy policy and notices to explain how you use AI, and label any public-facing chatbot as AI. No policy yet? Start with our guide to whether your website needs a privacy policy.
- Keep reviewing. The OAIC says due diligence for AI products shouldn’t be a “set and forget” approach.
Who does this bind? The Privacy Act covers organisations with an annual turnover above $3 million. It also covers some smaller businesses whatever their turnover, including private health service providers (allied health practitioners among them), gyms and weight loss clinics, child care centres, and businesses that buy or sell personal information. The rest are on the OAIC’s list.
Under the threshold and not on that list? The Act may not apply to you, but your customers will still expect their details to stay out of an AI company’s training data.
A hypothetical clinic and one pasted referral letter
Take a hypothetical physiotherapy clinic in Geelong. As an allied health provider it’s covered by the Privacy Act, whatever its turnover. The receptionist uses her own ChatGPT Plus account to tidy up emails, and one busy afternoon she pastes in a GP’s referral letter, with the patient’s name, date of birth and injury.
Walk it through:
- The plan. Plus is an individual plan, and data sharing is on by default. Unless she has switched it off, that chat can be used for training.
- The feedback. If she gives the draft a thumbs up, the whole conversation may go into training, even with the setting off.
- The purpose. The patient gave those details to get treatment. Pasting them into an AI tool is a disclosure to the company that runs it, and APP 6 asks whether that fits the purpose of collection or whether an exception applies.
- The sensitivity. Health information is sensitive information, the exact category the OAIC singles out to keep out.
Swap her to a Claude Pro account and the steps barely change. The fix is a business workspace with training off by default, a rule that patient details never go in, feedback buttons switched off for the team, and a privacy policy that says how the clinic uses AI.
What can your staff paste? A traffic-light rule
Most of the risk sits in what people paste on a busy day.
Light | What’s in the prompt | Rule |
|---|---|---|
Green | Public material: website copy, published posts, ad drafts, general research | Any plan is fine |
Amber | Confidential but not personal: pricing plans, financials, tenders, internal process documents | A business plan, or a personal plan with training switched off and no feedback ratings |
Red | Personal information about customers, staff or patients, above all health, financial or identity details | Keep it out of public AI tools, as the OAIC recommends. Consider a business workspace only after due diligence and a privacy policy update |
Make it part of onboarding. The OAIC notes that any rule against putting personal information into AI prompts needs training and auditing behind it.
How do you turn off training in ChatGPT and Claude?
Each is four steps. Do it on every staff account, on every device someone signs in from.
ChatGPT (web, signed in)
- Open your account menu.
- Select Settings.
- Select Data controls.
- Select Improve the model for everyone, turn it off, and select Done.
On the iOS and Android apps it’s the same path from your profile icon. Check voice mode’s own training control while you’re there, and Codex’s Include environments setting if anyone uses Codex.
Claude (desktop or browser)
- Select your name to open the settings menu.
- Select Settings.
- Select Privacy.
- Under Help improve our AI models, switch the toggle off.
The mobile app follows the same Settings, then Privacy, path.
For a one-off chat you don’t want used for training, use ChatGPT’s temporary chat or Claude’s incognito chat. Neither turns a red-light prompt green, though.
When should features decide it?
Once the plan and the setting are right, pick on the work. In the ChatGPT vs Claude features debate, the common view among regular users is that Claude is stronger on long-form writing and code, and ChatGPT is the broader all-rounder, with image generation and voice built in. That’s opinion, and it shifts with every release.
Run a one-week test instead:
- Pick five real tasks: a quote email, a social caption, a policy summary, a spreadsheet formula and one job you dread.
- Give both tools the same prompt, with no personal details in it.
- Score each answer on how much you had to fix.
- Check Claude pricing and ChatGPT pricing on each company’s own page before you buy, because tiers and inclusions change often.
Our ChatGPT for business breakdown covers day-to-day uses, and our list of AI tools worth paying for covers the rest of the budget.
There’s a fair argument the other way. If you never type anything personal or confidential, the privacy setting matters less and features are the whole decision. A sole trader drafting social captions can pick on taste. The plan question only bites once customer data or confidential numbers enter the chat, and that’s easy to do by accident on a busy day.
The bottom line
Pick the plan and the setting before you pick the bot
For an Australian business, Claude vs ChatGPT is a privacy decision before it’s a features one. On personal plans both can train on your chats while the setting is on, and the feedback buttons can send a whole conversation in anyway. Business plans keep your data out of training by default. Put the team on the right plan, switch off training and ratings, keep personal information out as the OAIC recommends, then choose the tool that does your work better.
Want AI working in your business without the data worry?
We build websites, AI agents and custom software for Australian businesses. See our custom software development services.
Important FAQs
Does ChatGPT train on my business data?
On ChatGPT Free, Plus and Pro, data sharing is on by default, so your chats can be used for training until you turn off Improve the model for everyone in Data controls. On ChatGPT Business, Enterprise and Edu, and on the API, OpenAI says it doesn’t use your inputs or outputs for training by default.
Does Claude use my chats for training?
On Claude Free, Pro and Max, Anthropic uses your chats for training only if you allow it, or if a chat is flagged for safety review. If you allow it, chats are kept for five years instead of 30 days. On Team, Enterprise and the API, Anthropic says it doesn’t train on your data by default.
Is Claude or ChatGPT better for a small business?
Both are capable, and many owners pay for both. The common view is that Claude is stronger for writing and code and ChatGPT is the broader all-rounder. For a business, get the plan and privacy setting right first, then test both on five real tasks and keep the one you fix least.
Can I put customer information into ChatGPT or Claude?
The OAIC’s best-practice advice is not to enter personal information, especially sensitive information, into publicly available generative AI tools. If your team needs AI for customer work, use a business workspace, do due diligence, check APP 6 for each use, and update your privacy policy to explain it.
