When you’re the whole company, an AI assistant can feel like a colleague. You paste in a customer email and ask for a kinder reply. You drop in a spreadsheet of sales and ask what stands out. It’s fast, and that speed makes it easy to forget a basic question: where does this text go after you hit enter?
The answer depends on the tool, your plan and your settings. Below is a practical way to think about it, plus what three popular consumer assistants say in their own help pages as of October 2026. Policies change, so treat this as a starting point and check the official pages yourself.
The Simple Rule: Don’t Paste What You Couldn’t Explain
A useful test: if you couldn’t comfortably explain to a customer that their details were pasted into a third-party AI tool, don’t paste them. That rules out a lot.
Keep these out of consumer AI chats unless you’ve checked the tool’s terms and settings and you’re confident it’s appropriate:
- Passwords, API keys, login codes and recovery codes
- Payment details, bank information and government ID numbers
- Customers’ names, email addresses and order histories
- Private messages from customers, collaborators or clients
- Unreleased product files you’d be upset to see elsewhere
- Anything covered by a contract or a confidentiality agreement
Google’s Gemini privacy page says this directly: don’t enter confidential information you wouldn’t want a reviewer to see or Google to use to improve its services, and don’t enter login or payment information.
Strip Details Before You Ask
Most of the time, the AI doesn’t need the private parts to help. Replace names with “Customer A,” remove email addresses and order numbers, and summarize instead of pasting whole threads. For spreadsheet questions, delete identifying columns first or use made-up sample rows that mirror the structure. You still get the help; the tool never sees the details.
What the Big Three Say About Your Chats
Here’s a quick summary of the official help pages for the consumer versions of three widely used assistants. Business and enterprise plans often have different terms.
ChatGPT (OpenAI)
OpenAI’s Data Controls FAQ says you can stop your conversations being used to train its models by turning off the “Improve the model for everyone” setting under Settings, then Data controls. It also offers Temporary Chats, which don’t appear in your history, don’t create memories and aren’t used to improve models, though OpenAI says they may be kept for up to 30 days for safety purposes.
Claude (Anthropic)
Anthropic’s privacy center says chats on consumer plans may be used to improve its models if you allow it in your settings (the Model Improvement setting), if a conversation is flagged for safety review, or if you explicitly opt in to a program. Incognito chats aren’t used for model improvement. Its retention page says deleted conversations are removed from back-end storage within 30 days, while chats used for training with the setting on may be kept in de-identified form for up to five years, and feedback you submit, like a thumbs up or down, is kept for five years.
Gemini (Google)
Google’s Gemini Apps privacy hub says that with Keep Activity on, your chats are saved and used to improve Google services, including training generative AI models, and that a subset of chats is reviewed by human reviewers after being disconnected from your account. With Keep Activity off, future chats aren’t used for training unless you send feedback, but Google still keeps them for 72 hours. Chats that human reviewers have already seen are kept for up to three years, even if you delete your activity.
Notice the pattern: each service gives you some control, and each keeps some data for some period regardless. “Turned off training” doesn’t mean “never stored.”
A Ten-Minute Settings Check
Once per tool you use regularly:
- Open the privacy or data controls page and decide whether you want your chats used for model improvement.
- Learn how the tool’s temporary or incognito mode works, and use it for anything you’d rather not keep in your history.
- Check what happens to files you upload and to connected apps such as cloud drives or email.
- Delete old chats you no longer need, and note the retention period the help page lists.
- Turn on two-factor authentication for the account. A leaked password exposes your whole chat history.
Your Customers’ Data Is a Promise
If you sell digital products or run an email list, you likely have a privacy policy that tells people how you handle their information. Pasting their details into an outside tool may not fit what you told them. Privacy laws such as GDPR for EU residents and various US state laws can also apply. This is general information, not legal advice; if you handle a lot of customer data, it’s worth getting proper guidance for your situation.
The Honest Trade-Off
None of this means AI tools are unsafe to use. It means they’re services run by companies, with settings, retention periods and human review processes you should understand. A few minutes of setup and a habit of stripping details let you keep the speed without handing over things that aren’t yours to share.
Do This Today
Pick the AI assistant you use most, open its privacy or data controls settings, and make a deliberate choice about model training. Then write a three-line “never paste” list on a sticky note next to your screen: passwords and keys, customer details, private messages.
