TwiddleLab / Artificial Intelligence / AI and Us: Using AI Well
Privacy: What You Give Away
What you will learn
- What you type into an AI is sent to the company's servers. Copies can be stored, reviewed for safety, or used for training.
- Personal data points to you or someone else: name, school, address, phone, card numbers, codes. Some is sensitive: health, location, passwords.
- Data minimization: give the AI what the task needs, not who you are. Use placeholders.
- Check the app's settings: many let you switch off training on your chats.
How the lesson goes
Sara sends a chatbot a message, then closes the app. Where can her message still be?
Make a bet: tap every place you think it can end up. Then Check.
Watch where Sara's message goes when she presses Send.
This is her full message, nothing removed.
Clean up Sara's message before it's sent.
Tap each part that points to her or to someone else, so it becomes a placeholder. Keep what the task needs. Then Check.
The note is just as good with placeholders: Sara fills the real names in herself after. What's never sent can't be stored, reviewed or trained on.
What you just protected has a name.
Personal data: anything that points to a person: name, school, address, phone, face, voice. Some is sensitive: health, money, passwords.
Data minimization: share only what the task needs. Placeholders like [my name] keep the task and drop the person.
Training on user data: some services use chats to improve future models unless you switch it off.
Fine to share with a chatbot, or keep out?
Drag each one into a lane.
Two questions sort almost anything: does it point to a person, and does it open something? If neither, share it.
Clean sweep: three messages, one after another.
Hide what points to a person, keep what the task needs, then Send. +1 for each detail hidden, -1 for each one left in or each needed part hidden. Score 8 of 9.
Before you paste anything into an AI tool.
Put the four moves in order.
Each step stops a different leak: the first two keep personal data out of the message, the third decides what the company may keep.
Code: you paste your program into a chatbot to fix a bug. Which lines must come out first?
Drag each line into a lane.
The chatbot needs your logic, never your secrets. Swap them for placeholders: the bug fix works just as well.
Which message is safest to send to a chatbot?
A placeholder gives the AI everything the task needs without anyone's personal data.
Last move: switch on your privacy shield. Which setting lets your chats train future models?
Tap it.
Many apps put training behind a friendly name like "Improve the model". Switching it off is your choice, and the app still works.
Previous lesson: Deepfakes and Media Literacy · Next lesson: Learning with AI, Not from It · Part of the course: AI and Us: Using AI Well
Lessons are interactive and need JavaScript. Turn it on to play them.
