AI Safety for Seniors
What's All This Artificial Intelligence About? · October 2026

Start here.

Everything from class, in the order you need it. This page is safeai4seniors.com/start — all one word, with the number 4.

Oct 6 · Week 1
GeminiGet it in your hand
Done ✓
Oct 13 · Week 2
ChatGPTTrust it for words. Check it for facts.
Done ✓
Oct 20 · Week 3
ClaudeSafety is a skill, not a fear
This week
Oct 27 · Week 4
All threeThings your kids don't know AI can do yet

1 Open this week's tool: Claude

Open Claude in your browser claude.ai — sign in with an email address or your Google account. Claude needs a sign-in; there's no try-it-without-an-account. Get the Claude app Search “Claude” — the one by Anthropic. Same sign-in as the website.
iPhone / iPad — App Store 🤖 Android — Google Play ↩ Gemini ↩ ChatGPT

Don't create an account under a clock. No Claude account yet? This week's lab works in any of the three tools — open the one you already have. If you'd like tool number three in your pocket, make the account at home: it asks you the privacy question at sign-up, and you can say no.

2 Flip the privacy switch (once)

  1. At sign-up Claude asks whether your chats may be used to improve it. You can say no.
  2. To check or change it later: tap your initial, then Settings, then Privacy.
  3. Find Help improve our AI models. If it's on, tap it — Off. Done.
  4. It remembers now. Since August 25 Claude keeps notes across your chats by default. Settings → Capabilities → Memory lets you read, edit or pause them. A new chat has an incognito option — it stays out of memory and out of training.

Paths checked against Anthropic's own privacy and help pages on October 18, 2026 — menus move, so tell Robert if yours looks different. Gemini's and ChatGPT's switches are on their Tool Cards below, and on workbook Week 3, page 3.

3 The one rule — and this week's version of it

Cover your numbers. And when you check a suspicious message with AI: paste their words, never yours — not your name, your account number or your address. It never needs them to help you.

4 Same three moves, new tool

⌨️
Type
Ask it in plain English, like a friend.
🎤
Talk
Tap the microphone and just say it.
📷
Show
Tap + or the camera. Photograph it and ask.

Gemini, then ChatGPT, now Claude. Every AI tool you'll ever meet, you'll meet the same way. Next week you'll onboard the next one yourself.

5 Something feels urgent? Run the four moves

You don't have to spot every scam. You have to run a procedure whenever something feels urgent — a call, a text, a letter, a pop-up. Real or fake, the moves are the same.

1
Pause.
Urgency is the trick. A real emergency survives a two-minute callback. A fake one can't.
2
Call back on a number you already have.
The back of your card, your paper statement, your own contacts. Never the number in the message — and never a number an AI gives you.
3
Use your family safe word.
A voice can be cloned from a few seconds of video. A word you agreed in person can't be.
4
Never gift cards, wire transfers or crypto.
No real agency or company demands payment that way. Ever. The moment anyone says “gift card,” the conversation is over.
AI is good at “that smells wrong.” AI is dangerous at “here's who to call.”

6 Checking a message with AI

Copy only their words from the text or email — never anything about you — paste them into Claude (or Gemini, or ChatGPT) and ask:

“Is this likely a scam? What are the warning signs? What should I do?”It is very good at this — urgency, odd links, a name that's almost right.
“Which of my four moves applies here?”
“Now argue the other side — what would make this message legitimate?”
“What questions should I ask before I believe this?”

The trap: if you then ask it for the company's phone number, it will happily give you one. Maybe it's right. You can't tell, and neither can it. Never call a number an AI hands you. The back of your card can't be wrong.

7 The family safe word — set it up today

  1. Pick one person — a child, a sibling, a friend who'd get the call. (Then a second person this week.)
  2. Pick a word that isn't on your Facebook — not a pet, not a street, not a team.
  3. Text them or call them — in person or by phone, never by replying to a message you didn't expect:
    “Our family safe word is ____. If anyone ever calls or texts saying they're me and in trouble, ask for it before you do anything.”
  4. In the workbook, write their name only — never the word. Give them a wallet card.

8 A photo or video that might be fake

Ask where it came from

  • Who posted it first? Is a news source you trust reporting it?
  • Does it want you to act fast, pay or share? That's the tell that matters.
  • When in doubt, don't share it.

Search with the photo itself

  • In the Google app, tap the camera in the search bar (Google Lens) and point it at the picture — where else does it appear?
  • Or ask your AI: “What questions should I ask before I believe this photo?”

Spotting fakes by eye gets harder every month. Asking where something came from doesn't.

9 Claude: fun fact / fine print

Fun fact

  • The free plan lets you type, talk out loud, show it a photo, and it can look things up on the web.
  • It's the only one of the three that asks you at sign-up whether your chats may be used to train it.

Fine print

  • It needs a sign-in — no try-it-without-an-account. Free usage comes in five-hour windows; “come back later” is all a limit means.
  • Memory is on by default (since August 25). Settings → Capabilities → Memory to see or pause it; incognito chats stay out of both memory and training.

Checked October 18, 2026 against Anthropic's own pricing, privacy and release-notes pages. The three moves don't change; the fine print does — that's why you check it.

10 This week's materials

📄 Claude Tool CardOne page: open it, the privacy switch, memory and incognito, the three moves, the one rule. On your chair Tuesday — and here now. 📄 Four-Moves wallet cardEight to a page — print, cut, keep one in your wallet and give one to the person you agree the safe word with.

Week 3 slides and workbook pages are posted here after class on October 20.

This week's homework (workbook Week 3, page 3): have the safe-word conversation with a second person (names on the page, never the word). Check the privacy switch on every AI app on your phone — page 3 has all three paths. And tell one friend the four moves — teaching it is how you keep it.

Last week (ChatGPT):

📄 Week 2 slidesThe second question, words vs. facts, the demo, the lab cards. 📄 Week 2 take-home: the second question, words vs. facts, the fine printEvery fact dated and sourced. Two pages. 📄 ChatGPT Tool CardOne page: open it, the privacy switch, the three moves, the one rule. 📄 Week 2 workbook packetThe follow-ups, words vs. facts, the Notebook steps, the four cards, homework.

Week 1 (Gemini): slides · take-home · Gemini Tool Card · workbook packet — or see all class materials.

11 Next Tuesday — the finale

Things your kids don't know AI can do yet All three tools on the table. In pairs you'll build something real — an app you describe into existence, a podcast made from a document, a photo brought back to life — and show the room. Bring one thing you'd love to make, learn or finally understand.
All class materials This cohort's pages as they're posted — and the August 2026 archive.

Useful AI, minimal exposure, human verification.

Safety is a skill, not a fear.

Same three moves, new tool.

AI is good at “that smells wrong.” AI is dangerous at “here's who to call.”

Questions between classes? Email Robert any time: rsellers@sonic.net.