Home / GUIDE

GONDAL AI / GUIDE

AI Privacy Basics

Every AI assistant collects your prompts. Most use them to improve their models unless you say otherwise. Here is what actually happens to your data — and what you can do about it.

Plain-language note: policies change. What follows reflects the state of things as of October 2026, drawn from the providers' published policies and independent audits. When in doubt, check the provider's current privacy page.

What assistants typically collect

• Your prompts and the responses — every word, including corrections and follow-ups.
• Files you upload — documents, images, spreadsheets, code.
• Account and device metadata — email, IP address, location, device type, usage patterns.
• Integration data — if the assistant can read your email or calendar to help you, it can see that content too. Convenient and worth understanding.

The training question

The key question: are my chats used to train future models? As of 2026, the default answer on the major consumer plans is generally yes, unless you opt out — though defaults and exceptions (including regional differences) have shifted over time, so verify.

• ChatGPT: Settings → Data Controls → turn off the model-improvement toggle. Temporary chats are excluded from training regardless.
• Claude: Settings → Privacy → turn off the improvement toggle. API traffic is excluded from training by default.
• Gemini: turn off “Gemini Apps Activity” in your Google account's data and privacy settings; you can also delete past activity there.
• Copilot: check the privacy toggle in settings; business and enterprise tiers exclude content from training by contract.

Two things most people miss: human reviewers may read sampled conversations at all major providers for safety and quality — and shared conversation links expose the entire thread to anyone with the link.

Practical tips

• Never paste secrets. Passwords, API keys, financial credentials, and other people's private data do not belong in any chatbot.
• Use temporary or incognito chats for sensitive one-off questions.
• Opt out of training on every assistant you use regularly — it does not reduce answer quality.
• For client work (legal, health, finance), the business tier of your chosen platform is worth it: training exclusion is contractual, not just a toggle.
• Check retention: know how long deleted chats are kept and what survives deletion for safety monitoring.

What to check in any AI privacy policy

• Is my content used for training, and what is the default?
• Can I opt out, and is it easy to find?
• How long is data retained, including after I delete it?
• Do humans review conversations, and under what conditions?
• What happens to my data if I cancel?

If a provider cannot answer those five questions clearly, that is itself an answer.