theaisafemode.com
AI Safe Mode Guide

What Not to Share With AI Chatbots: A Simple Privacy Checklist

A Stanford study found six leading US AI developers use chat inputs to train their models by default. Here's what to keep out of your chats and how to limit the risk.

AI chatbots feel private. It’s just you and a text box. But what you type may be stored, reviewed and used to train future models.

In October 2025, researchers at the Stanford Institute for Human-Centered AI (HAI) compared the privacy policies of six leading US AI developers. They found all six use users’ chat data by default to train or improve their models.[1] This guide turns that finding into a simple checklist.

What the research found

The Stanford team analysed 28 policy documents across the six developers. According to Stanford HAI:[1]

  • all six use chat inputs by default to train or improve their models
  • some keep this data in their systems indefinitely
  • some, but not all, say they de-identify personal information before training
  • some allow humans to review chat transcripts for training
  • at companies with many products, chats can be combined with data from search, shopping and social media

The study’s lead author warned that sensitive information shared in a chat may be collected and used for training, even if it is in a file you uploaded during the conversation.[1]

Regulators are paying attention too. In September 2025 the US Federal Trade Commission ordered seven companies that run consumer AI chatbots to explain, among other things, how they use or share personal information from users’ conversations.[2]

The “never paste” list

Treat an AI chat like a message that could one day be read by someone else. Keep these out:

  1. Passwords, 2FA codes and recovery codes. No chatbot needs them.
  2. Card numbers, bank details and ID numbers such as passport or national ID.
  3. Health details tied to your name, including test results and prescriptions.
  4. Other people’s private information: a friend’s messages, a colleague’s file, a child’s photo.
  5. Work secrets: customer data, unreleased plans, source code you’re not allowed to share, and API keys.
  6. Exact location and routine, such as your home address plus when you’re away.

If you need help with a document, remove names, numbers and addresses first, or describe the problem in general terms instead.

How to stay safe

  • Turn off training on your chats where the service lets you. The Stanford researchers advise opting out whenever possible.[1]
  • Redact before you paste. Replace names and numbers with placeholders like [NAME] and [ACCOUNT].
  • Be careful with shared links. A shared chat link can be opened by anyone who has it.
  • Check the retention settings and delete chats you no longer need, knowing that deletion may not be instant.[1]
  • Watch what children share. Companion-style chatbots are designed to feel like a friend, which can make sharing feel safe.[2]

Why this matters even if you “have nothing to hide”

Leaks happen. AI Safe Mode has covered an AI chat app flaw that could expose a whole chat history, an AI robot toy leak that exposed private chats, and a warning that children’s AI chats could be sold in a bankruptcy. Data that is never typed can’t leak.

For families, read our guide on AI companion chatbots and kids.

Protect yourself with Privaku

Sources

1. Stanford HAI (Oct 15, 2025): “Be Careful What You Tell Your AI Chatbot”. https://hai.stanford.edu/news/be-careful-what-you-tell-your-ai-chatbot 2. US Federal Trade Commission (Sep 11, 2025): “FTC Launches Inquiry into AI Chatbots Acting as Companions”. https://www.ftc.gov/news-events/news/press-releases/2025/09/ftc-launches-inquiry-ai-chatbots-acting-companions

FAQ

Do AI chatbots use my conversations for training?

Often, yes. A Stanford study of six leading US AI developers' privacy policies, published in October 2025, found all six use users' chat data by default to train or improve their models. Some let you opt out.

Can a human read my AI chats?

Possibly. The same Stanford study found that some developers allow humans to review users' chat transcripts for model training.

Does deleting a chat remove it everywhere?

Not necessarily. The Stanford researchers found some developers keep chat data in their systems indefinitely, so check each service's retention policy and settings.

See the post on YouTube Watch and follow AI Safe Mode on YouTube @AISafeMode and Instagram @aisafemode.

References

  1. Stanford HAI, Oct 15, 2025. Be Careful What You Tell Your AI Chatbot. Read the report ↗
  2. US Federal Trade Commission, Sep 11, 2025. FTC Launches Inquiry into AI Chatbots Acting as Companions. Read the report ↗
Facts come only from the published reports above and are attributed to them. AI Safe Mode doesn't name companies, products or people involved in incidents. Spotted an error? Email [email protected] and it will be reviewed and corrected. See the corrections policy.