Talking to AI
All lessons

Lesson 6/84th D · Diligence 9 min Last updated:

Diligence: privacy, honesty and responsibility

What never to type into an AI assistant, when to say you used it, and why the last word is always yours.

The last D may be the most important: you are responsible for how you use AI and for what comes out of it. The framework splits it into three moments.

1. When you choose (what to use and what to share)

Protect your data

When you type in a chat, the text is sent to the servers of the company offering the service. Depending on the settings, conversations may be stored, read by reviewers or used to improve the model.

Never type into a chat

Passwords and access codes · full credit card numbers and bank account details · PINs and one-time codes · ID documents · your or other people’s health data without thinking · confidential work information.

The good news: you can almost always get the same help without personal data. Just replace it with placeholders like [NAME] or [ADDRESS].

Privacy exercise

Make the text anonymous

You want the AI to help you answer this email. Click the personal information to hide it, then press “Check”.

Dear ,
I’m writing on behalf of my mother, , born on , patient ID .
After her visit on she was prescribed a new treatment for , but she has had and we’d like to know whether to continue.
Could you call us back on ? We live at .
Thanks,

Check your settings

Almost every assistant has an option to not use your chats for training and a temporary-chat mode. Look under “Settings” → “Privacy” or “Data”. If you use AI for work, ask which tools your company has approved.

Game: sort the cards

Would you type it in a chat?

For each item decide: fine, better anonymised, or never.

2. When you share (transparency)

If you used AI in a significant way for a piece of work, say so when it matters: at school, university, in an article, in a work document. Nothing to be ashamed of: it’s a tool, like a calculator or spell-checker.

  • At school and work, check the rules: some allow AI, some don’t, some only in certain ways.
  • A simple formula: “I used an AI assistant for the first draft and proofreading; the content and fact-checking are mine.”
  • Don’t pass off as your own a creative text written entirely by AI, if the person receiving it expects you wrote it.

3. When you publish or hand in (responsibility)

Before sending, signing or publishing something made with AI:

  • Read all of it. Not just the beginning.
  • Check facts, numbers and names (Discernment lesson).
  • Ask yourself: “Would I be comfortable defending every sentence of this text?”

The golden rule

AI is a collaborator, not a scapegoat. If you sign it, it’s yours: mistakes included.

Other risks to know about

  • AI scams: cloned voices, fake photos, flawless emails. If a “relative” calls asking for urgent money, hang up and call back on the number you know.
  • Fake images and videos (deepfakes): before sharing something shocking, look for confirmation from reliable sources.
  • Your relationship with AI: it’s a great tool for venting or reflecting, but it doesn’t replace friends, family or a mental health professional.

Quiz

1/3 You want help writing a complaint to your bank. What do you paste into the chat?