Using AI safely and honestly
These tools are useful, but how you use them matters. This Lesson covers a few habits. Keep private things out of the chat. Check facts before you trust them. Say when you used AI. Watch for bias, opt out of training if you want, and treat the tool as a tool, not an authority.
Do not paste in secrets or other people’s private data
Whatever you type into a chatbot leaves your device and goes to a company’s servers. On a free plan, it may also be stored and used to improve the model, and staff may review some chats.
So keep these out of the box:
- Passwords, bank or card numbers, and one-time codes.
- Your full ID numbers — passport, national ID, social security.
- Anything private about someone else: a friend’s health, a coworker’s contract, a client’s records.
- Work documents your employer treats as confidential.
You can still get help. Describe the situation in general words. Swap real names for “my manager” or “the client.” The tool does not need the real details to fix your tone or structure.
Verify anything factual before you rely on it
An earlier Lesson showed that these tools can state a wrong fact in a confident voice. That does not stop mattering when the answer sounds careful.
Some facts need a real source before you act on them. A date, a price, a legal rule, a medical dose, a citation. Confirm each one against a search tool that links to pages, or a person whose job it is to know.
Say when you used AI, where it matters
For a private text or your own to-do list, how you wrote it is your business. It is different when someone is judging the work as yours:
- School: many schools now have clear rules. Some allow AI help if you disclose it; some do not allow it at all. Ask before you use it.
- Work: your employer may have a policy on AI-written material, especially anything sent to customers or the public.
- Publishing: if you put something out under your name — an article, a review, a photo — readers are trusting it is yours. Disclose AI use when it would change how they read it.
The simple test: if someone would feel misled to learn AI helped, say so up front.
Its answers carry the bias of its training text
The model learned from a huge amount of human writing, and that writing is not evenly spread. Some places, languages, and viewpoints are far more represented than others.
So its “normal” answer leans toward whatever was most common in that text. Ask it for a picture of a “doctor” or a list of “great novels” and the lean shows. It is not neutral, and it does not know that. Treat its default as one common view, not the only one.
How to keep your chats out of training
Most major tools now have a setting to keep your conversations out of model training. The exact path changes, so check the app. As of late 2026:
- ChatGPT: in Settings, open Data Controls and turn off the option about improving the model for everyone.
- Gemini: in Settings, turn off “Gemini Apps Activity” (also at myactivity.google.com).
- Claude: in Settings, open Privacy and turn off the option about helping improve Claude.
- Copilot (Microsoft): in Settings, look under Privacy for a model-training toggle and turn it off.
- Grok: in Settings, look for a data-sharing option about training and turn it off.
Turning chat history off usually drops it from training too. The switch helps, but it is a backstop. The main habit is still to keep sensitive things out of the chat.
It is a tool, not an authority
A chatbot has no job title, no license, and no accountability. It cannot be your doctor, lawyer, accountant, or therapist. It can help you understand a topic, draft a question for a real professional, or organize what you already know.
For decisions about your health, money, legal standing, or safety, a person or an official source has to be the one you rely on.
Try it: run the safety check on a real task
Pick something you would actually want help with this week. Before you open a chatbot, run it through this check.
Strip out what is private, check what is factual, and be honest about AI’s part when the work is judged as yours. That is the whole habit.
Next, the last Lesson puts all of this to work on real everyday tasks.