Is it safe to put client data into ChatGPT or Claude? Data privacy for Australian businesses (2026)

Yes, for most everyday client work, as long as your team works on a business plan (ChatGPT Business or Enterprise, Claude Team or Enterprise) and everyone knows what stays out. On free or personal accounts, keep client personal information out altogether, which is what Australia's privacy regulator, the OAIC, recommends as best practice for publicly available AI tools.

Below is what each plan does with your data, where it's stored, what the OAIC expects and the rules we set with clients. We checked every vendor fact on the vendors' own pages on 16 November 2026, and these policies change often, so check again before you rely on them.

Personal plans

ChatGPT Free, Plus, Pro. Claude Free, Pro, Max, Ultra.

  • Chats can be used to improve the vendor's models unless each person changes a setting
  • Each person controls their own chats

Business plans

ChatGPT Business and Enterprise. Claude Team and Enterprise.

  • Not used for training by default
  • Someone in the business controls the workspace

The exception: on business plans, a thumbs up or down, or a bug report, can send that conversation for training. Don't use the feedback buttons on chats that contain client details.

The biggest difference between the two. The full comparison is in the table below.

Personal accounts and business plans handle data differently

The biggest difference is model training. On personal plans, your chats can be used to improve the vendor's models unless the person changes a setting. On business plans they aren't used by default, and someone in the business controls the workspace.

PlanUsed to train the vendor's modelsWho controls how long chats are keptData stored in Australia
ChatGPT Free, Plus or ProYes, unless each person turns off "Improve the model for everyone"Each personNot available
ChatGPT BusinessNot by defaultWorkspace admins. Deleted chats leave OpenAI's systems within 30 daysRolling out to some new workspaces, with a copy still kept in the US for safety checks
ChatGPT EnterpriseNot by defaultWorkspace admins. Deleted chats leave OpenAI's systems within 30 daysAvailable for new workspaces
Claude Free, Pro, Max or UltraOnly if the person allows it in the "Model Improvement" settingEach personNo, stored in the US
Claude TeamNot by defaultEach person deletes their own chats. Deleted chats leave Anthropic's systems within 30 daysNo, stored in the US
Claude EnterpriseNot by defaultAdmins can set a retention period, 30 days at the shortestNo, stored in the US

Both vendors make an exception on business plans for feedback. If someone clicks thumbs up or down, or reports a bug, that conversation can be used for training. Anthropic keeps feedback for up to five years. Our rule of thumb: don't use the feedback buttons on chats that contain client details. On Claude Team and Enterprise, owners can switch the buttons off for everyone.

Anthropic also keeps chats flagged for breaching its usage policy for up to two years, even on Team, though everyday client work is unlikely to trigger that.

Free tool AI Use Policy Template: a short, plain-English policy your team can read in five minutes. Fill in the bits in [brackets] and share it.

What Australia's privacy regulator says

The Office of the Australian Information Commissioner published guidance on using commercially available AI products in October 2024 and updated it in January 2025. It covers paid business tools as well as free chatbots. The parts that matter most for a smaller business:

  • Keep personal information out of public tools. As a matter of best practice, the OAIC recommends you don't enter personal information, and particularly sensitive information, into publicly available generative AI tools.
  • Your privacy obligations follow the data. They apply to what goes into the tool and to anything it produces that contains personal information.
  • Do your homework on the product. The OAIC asks you to check whether the developer or other third parties will have access to the data you put in or generate. The table above answers that for ChatGPT and Claude.
  • Say how you use AI. Your privacy policy should include clear information about your use of AI.
  • Use information for the purpose you collected it. Using it for a different purpose needs consent or has to be something the person would reasonably expect, which the OAIC says will often be hard to show for AI uses.
  • Know where it goes. The OAIC points out that cloud AI tools may send personal information outside Australia.

The Privacy Act covers businesses with annual turnover over $3 million, plus some smaller ones whatever their turnover, including health service providers and businesses that trade in personal information. The exposure draft privacy bill released on 31 August 2026 doesn't remove that small business exemption, but it's still a draft.

If you're under the threshold, your client contracts and confidentiality duties still apply, and so do any professional rules in your industry.

If you'd like your settings, policy and access checked as part of a proper setup, that's what AI implementation covers.

Where your data is stored

OpenAI offers Australian data storage for new ChatGPT Enterprise and Edu workspaces and eligible API customers. On those workspaces, conversations, files, memory and other saved content are stored at rest in Australia, while workspace details, billing information, logins and anything sent to connected apps or web search can sit outside the region.

The model processing itself can only be kept in-region in Europe, the US and the UAE for now.

ChatGPT Business is starting to offer a choice of storage region at checkout, but it isn't available to everyone yet, and a copy of every prompt and response is still kept in the US for a limited time for safety checks.

Anthropic says Claude data is stored in the US, and that it may route traffic through the US, Europe, Asia and Australia. In March 2026 it said it was exploring more computing capacity in Australia, but until its privacy pages say otherwise, plan on Claude data being stored in the US.

If a client contract or your industry requires data to stay in Australia, Claude Team doesn't meet that today, and ChatGPT Business only partly does, because copies are still kept in the US for safety checks. Read the contract wording, and ask the client if it's unclear.

Our rules for client information

This is the starting point we use with clients. Adjust it for your industry and your contracts.

What it isOn a business planOn a personal or free account
Drafting, rewriting or summarising with no client detailsFineFine
Client names and the details a task needsFine, but include only what the task needsNo
Health, financial or other sensitive informationOnly once you've checked your obligations and what clients expectNo
Passwords, account numbers and tax file numbersNoNo
Anything a client asked you to keep confidentialCheck the engagement terms firstNo

Five settings to check before anyone pastes in client data

  1. One business workspace, on work accounts. Turn on single sign-on if you have Microsoft 365 or Google Workspace, and move anyone doing client work off personal accounts.
  2. Training and feedback. Confirm the workspace isn't sharing data for training. On Claude Team and Enterprise, owners can turn off the feedback buttons for everyone with the "Rate chats" setting. On ChatGPT, tell people what the feedback buttons do.
  3. How long chats are kept. On ChatGPT Business and Enterprise, admins set this. On Claude Enterprise, admins can set a custom period. On Claude Team, people delete their own chats.
  4. Connectors, one at a time. OpenAI says information sent to a connected app follows that provider's own terms, so connect one source, check what it can see, then add the next.
  5. Who can read conversations. On ChatGPT Business, workspace admins can view, export and delete members' conversations. On Claude Team, the Primary Owner can export the organisation's conversation data. That matters for HR and partner-only matters, so choose your account owner with care.

Then write it down. Our free AI use policy template covers approved tools, what can go in and how to check work, and takes about 20 minutes to fill in.

Common questions

Want a hand?

If you'd like the settings checked and your team trained on what's safe to put in, that's part of Claude for Teams and ChatGPT for Teams. Training starts from $3,500 + GST and the full setup from $7,500 + GST, and you buy the seats directly from Anthropic or OpenAI.