Is it safe? Privacy basics everyone should know

The honest answer

AI tools are not perfectly private. When you type something into ChatGPT, Claude, or Gemini, that text goes to a server owned by a large technology company. How long it’s stored, whether it’s used to train future models, and who can see it depends on the tool and your account settings.

That said — for everyday use, AI tools are not meaningfully more risky than using email or Google Search. The risks are real but specific. Knowing them lets you use these tools confidently.

What actually happens to your conversations

Most major AI companies — OpenAI, Anthropic, Google — use conversation data to improve their models by default. You can usually opt out in settings. Enterprise and paid business plans typically have stronger protections and explicit commitments not to use your data for training.

The practical reality: your conversation about brainstorming birthday gift ideas is not sensitive. Your conversation containing your company’s unreleased financial projections is.

What you should never type into a free AI tool

Specific things to keep out of AI conversations:

Passwords, API keys, or login credentials of any kind. Full names combined with addresses, dates of birth, or ID numbers. Confidential business information — unreleased product plans, financial data, client lists. Medical information that you wouldn’t want associated with your name. Legal matters where confidentiality is important.

How to use AI more privately

Turn off chat history when available — most tools have this option in settings. It usually means the conversation isn’t used for training and isn’t stored long-term. In ChatGPT this is called “Temporary chat.” In Claude it’s available in settings.

Use a professional or business account if your company provides one — these typically have stricter data handling agreements.

For sensitive work, consider self-hosted or enterprise AI options where your data doesn’t leave your organisation’s systems.

The real risk level for most people

For the vast majority of everyday use — writing help, explanations, brainstorming, summarising public documents — the privacy risk of using mainstream AI tools is low. The companies involved have strong reputations to protect and explicit policies against selling your data to advertisers.

The risk becomes meaningful when you start pasting genuinely sensitive information. Know the line, and you can use these tools freely on the right side of it.