tools.ceo

Is Claude AI safe for business data? Privacy, training and controls

Claude AI safe for business data? It depends on your plan. What Anthropic says about training, retention and admin controls, and what Australian owners should check.

Keeping up with AIUpdated November 3, 20264 min readBy tools.ceo editorial

Padlock on a keyboard representing data security
Photo: Unsplash

The short answer

Claude AI can be used safely with business data if you choose the right plan and set clear rules. Anthropic lists Team and Enterprise plans as not training on your content by default, while Free, Pro and Max have a training setting you control. No plan removes your own duties under the Privacy Act. Before staff paste in client information, pick a business plan, switch off optional sharing, restrict connectors, and write a one-page rule about what must never go into any AI tool.

  • Team and Enterprise are listed as not training on your content by default.
  • Individual plans have a training setting you must check and switch off if you want it off.
  • Your Privacy Act duties stay with you, whatever the AI vendor promises.
  • Staff rules and connector controls matter as much as the vendor’s policies.

Safe depends on the plan

There is no single answer because the plan sets the terms. Anthropic’s pricing page lists model training as opt-out for Free, Pro and Max, and as none by default for Team and Enterprise. Read the privacy documentation for your plan, because terms are updated and a summary like this one can go out of date. Take a screenshot or copy of the terms on the day you sign up, so you can show what you relied on later. Sources: Claude pricing, Anthropic Privacy Center

What training and retention mean

Training means your conversations could be used to improve future models. Retention means how long conversations are stored. These are separate settings. On individual plans Anthropic gives a control for whether your chats help improve Claude. Check the setting yourself rather than assuming, and note that safety review exceptions may apply. Ask your plan administrator or Anthropic directly if you need retention in writing. Sources: Anthropic Privacy Center

Admin controls on Team and Enterprise

Anthropic lists central billing and admin controls on Team, with audit logs, role-based access, SCIM, a Compliance API and custom retention controls on Enterprise. Owners decide which connectors staff may use. If you have to prove who accessed what, Enterprise is the plan to ask about. Team gives you most of what a small business needs, but it lacks some of the audit features listed for Enterprise. Sources: Claude pricing, Claude docs: get started with connectors

Your Australian obligations

The Office of the Australian Information Commissioner advises businesses that using AI products does not remove their Privacy Act obligations, and that personal information should not be entered into public AI tools unless it is clearly necessary. Read the OAIC guidance, and ask your lawyer about notification and cross-border disclosure for your situation. Your lawyer can also tell you whether any disclosure to a customer is required when their information is processed by an AI tool. Sources: OAIC: guidance on privacy and the use of commercially available AI products

Connectors widen the risk

A connector lets Claude read or act inside a service, such as your drive or inbox. Anthropic lets you set tool permissions to always allow, needs approval or blocked. Start with read-only access, require approval for actions that send or change things, and review who has connected what. Review connections quarterly and remove those that nobody uses, and ask staff to tell you before they connect anything new. Sources: Claude docs: get started with connectors

A one-page staff rule

Write down what may and may not be pasted into AI: no client identifiers, health or financial details, passwords or unreleased deal information. Say which plan is approved and that personal accounts are not. Review the rule every six months, because tools and terms change faster than policies do. Keep it short enough that people will actually read it, and run a ten-minute briefing when it goes out.

Questions people ask

Does Claude train on my data?

It depends on the plan. Anthropic lists Team and Enterprise as none by default, and Free, Pro and Max as an opt-out setting you control. Check the current settings.

Is Claude compliant with Australian privacy law?

Compliance is your responsibility as the business. Use the OAIC guidance and speak to a lawyer about your use.

Can I use Claude with client documents?

Only on a plan and with settings your risk review supports, and only when clearly necessary. Strip identifying details where you can.

tools.ceo is published by Kamand, which is mentioned on this page. No affiliate links on this page at publication. Product names link to official sites so you can check current pricing. How we work.

Keep reading.

New guides, about once a fortnight.

Websites, security and AI for people who pay for the software. One email when something is worth your time, nothing else.

Free. Unsubscribe from any email.