Insights · AI and security

Is it safe to put company data into ChatGPT?

November 2026 · Fusion Technology Group

Short answer: it depends on which account you use and what you put in. A personal or free account is the wrong place for company information. A business plan is far safer, but it doesn’t remove the need for rules.

The account matters more than the brand

The same tool can treat your information very differently depending on the plan. At the time of writing:

ServicePersonal plansBusiness plans
ChatGPTFree, Plus and Pro content may be used for training unless you turn it offBusiness, Enterprise and API aren’t used for training by default
ClaudeFree, Pro and Max depend on the user’s model-training settingCommercial products aren’t used for training by default
Microsoft CopilotPersonal Microsoft accounts have consumer termsWork accounts have enterprise data protection

Terms change, so check the current wording for the plan you’re actually on.

“Not used for training” isn’t the same as private

A business plan solves one problem. These remain:

  • Your information is still stored for a period, usually outside South Africa, which has POPIA implications.
  • Accounts get compromised. A stolen login can expose a person’s entire chat history.
  • AI can find what it shouldn’t. Connected to your files, it can surface a payroll sheet shared too widely years ago.
  • Documents can carry hidden instructions. An email or PDF can contain text telling the AI to do something you didn’t ask for.
  • You may already have promised not to. NDAs and client contracts can restrict where their information goes.

This isn’t theoretical

In 2023, Samsung engineers pasted source code and internal meeting notes into ChatGPT within weeks of being allowed to use it, and the company restricted the tool. IBM’s 2025 breach research found one in five organisations reported a breach involving unsanctioned AI use. These people weren’t being careless on purpose. They were trying to get their work done, and nobody had told them where the line was.

A simple rule for your team

Usually fineBusiness plan only, within your roleNever in a chat
Rewriting an email, explaining a formula, drafting a procedure, public informationCustomer orders, quotes, margins, supplier terms, contractsPasswords, ID numbers, bank details, salaries, medical or disciplinary records

A useful test: would you be comfortable seeing this in an audit log with your name next to it? If you’re unsure, leave it out or replace names with “Customer A”. Removing names helps, but small details can still identify someone.

If it’s already happened

Don’t panic, and don’t punish the person who told you. Establish what was shared and on which account. If personal information was involved, bring in your Information Officer to decide whether it needs to be reported under POPIA, and change any passwords or keys that were exposed.

The fix is straightforward

Give people one approved tool on a business plan, signed in with their work account. Tell them plainly what can and can’t go in. Clean up file sharing before you connect AI to your documents. Make it safe to report mistakes.

Want the full framework?

Our free working paper sets it out for South African owners and executives: what must never go into AI, who should approve what, how to connect AI to your systems safely, and a ten-point readiness check.

Get the working paper

Or take the 2-minute AI readiness check.

This article is general information and not legal advice. Vendor terms were checked in November 2026 and can change. Published November 2026.