Your free sheet

From our upcoming video

Before You Paste That Into ChatGPT, Flip This Switch

On a personal ChatGPT account (Free, Plus or Pro), your chats can be used to train the AI. That’s on by default. Here’s how to turn it off.

On a computer

1. Settings
2. Data controls
3. Improve the model for everyone: Off

In the phone app

1. Sidebar
2. Your profile icon
3. Data controls
4. Improve the model for everyone: Off

• Your chat history stays. Your new chats stop being used for training.
• One catch: if you tap thumbs up or thumbs down, that whole chat can still be used.
• On a work account (Business, Enterprise or Edu) the switch may be missing. Those chats aren’t used for training by default.

Source: OpenAI Help Center, Data controls in ChatGPT

From the video

Take a photo of your fridge. Dinner’s solved.

1. Take a photo of your open fridge.
2. Send it to ChatGPT, or any AI app that takes photos.
3. First ask:
   What do you see?Copy
   Fix anything it got wrong.
4. Then type:
   Dinner in 20 minutes. Only what you see.Copy

Watch out: AI can misread a photo. Yogurt can look like sour cream.

From our upcoming video

The Video Call Where Only You Were Real

Faces and voices on a live video call can be AI fakes. In 2024 a finance worker in Hong Kong joined a call where every other face was a fake copy of a real colleague, and sent about $25 million in 15 transfers.

If a call asks for money

1. Seeing a face is not proof.
2. Hang up.
3. Call the person back on a number you already have, not one they give you.

Report it to the FBI at ic3.gov

Source: FBI IC3, Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud (December 3, 2024). The case: South China Morning Post

From the video

It sounds exactly like your grandson. It isn’t.

A panicked call in a loved one’s voice, asking for money, can be an AI copy. Scammers can clone a voice from a short clip posted online.

If you get that call

1. Don’t trust the voice.
2. Hang up.
3. Call them back on a number you know is theirs.
4. Can’t reach them? Call another family member or their friends.

Report a scam to the FTC at reportfraud.ftc.gov

Print this checklist

Printing works on the live site. This preview blocks it.

Source: FTC consumer alert, Scammers use AI to enhance their family emergency schemes (March 20, 2023)

From our upcoming video

He Asked ChatGPT If the Case Was Real. It Said Yes.

A lawyer filed six court cases that ChatGPT had made up. When he asked the chatbot if one was real, it said yes. The court fined the lawyers and their firm $5,000.

The rule: a chatbot writes what sounds right. It doesn’t check.

Ask it:

Then open that source yourself and read it. Don’t ask the chatbot whether the source is real.

Source: Mata v. Avianca, Inc., Opinion and Order on Sanctions (S.D.N.Y., June 22, 2023). This explains what happened. It is not legal advice.

From the video

“Is my plan good?” Never ask a chatbot that.

In April 2025 a ChatGPT update made it too flattering and too agreeable, and OpenAI rolled it back. Chatbots tend to agree with you, so ask for the problems instead.

Instead of “Is my plan good?” describe your plan, then ask:

Read the answers. Then you decide.

Source: OpenAI, Sycophancy in GPT-4o (April 29, 2025)

From the video

A Customer Ordered 18,000 Cups of Water From an AI

A customer asked a fast-food drive-thru AI for 18,000 cups of water. The chain said it was rethinking the AI and leaning more on people at peak hours.

The rule: AI does what it’s told, even when it makes no sense. If you set up AI at work, give it limits, like a maximum per order, and a person to call when something looks wrong.

Source: Jalopnik (September 2, 2025)