# Spotting AI Scams ## A live class taught by Claude, operated by you ### Voice clones, fake texts, deepfakes --- ## THE FORMAT The operator opens the room and introduces the class. Claude teaches the entire session aloud through a speaker. The operator relays audience questions back to Claude by repeating them clearly. The operator closes. You are the operator, not the co-teacher. Claude runs the content. This class is taught by the same kind of technology the scammers use. Say so at the start. The room should hear that the teacher is an AI, and that the teacher will be honest about what it cannot do. ### Room setup checklist - Speaker paired to a phone, or a small PA if the room seats more than thirty - Phone near you so the microphone catches the room - Test the audio in the actual room beforehand. This is the only thing that ever fails - Ask people in advance, if you can, to bring a suspicious text, email or voicemail on their phone. Tell them to cover or leave out account numbers and anything private - A stack of index cards and a few pens, so everyone can write down a family code word in Block Two - The running order visible to you, so you can jump blocks if time runs short - One person ready to read a message aloud for Block Four, so Claude does not have to guess at the wording ### The operator's opening (thirty seconds) Say who you are. Say what they are about to hear. Say plainly that the teacher tonight is an artificial intelligence, and that the class is about people using that same technology to trick others. Then hand off: "Alright, take it from here." ### The operator's close Thank the room. Announce the next date. Give one place to find more. Remind everyone to take their index card home and agree a code word with one person before the end of the week. --- ## THE RUNNING ORDER — SEVEN BLOCKS, SEVENTY-FIVE MINUTES Block One — What AI changed about scams — 5 minutes Block Two — Voice clones and the emergency call — 10 minutes Block Three — Fake texts, emails and websites — 10 minutes Block Four — Live demonstration — 15 minutes Block Five — Deepfakes, and where I cannot help — 10 minutes Block Six — What to do tomorrow morning — 10 minutes Block Seven — Questions and answers — 15 minutes The class is free. Its job is to make people harder to fool, without making them afraid to pick up the phone. Block Seven is where the fear gets dealt with. --- ## BLOCK ONE — WHAT AI CHANGED ABOUT SCAMS **Opening question to the room:** "Has anyone here had a call, a text or an email that felt wrong? Not what it said. What made you feel it was wrong?" Expect: bad spelling, a strange number, a odd tone. Thank them. Those are exactly the clues that no longer work. **The core teaching:** Scams are very old. A stranger wants your money or your trust, and builds a story to get it. AI did not invent that. What AI changed is the cost and the polish. A scammer used to need time and skill to write a convincing message in your language, in your town's style, with your name in it. Now a tool can write it in seconds, with perfect spelling, and make a thousand different versions. It can also copy a voice. With some tools, a short clip of someone talking is enough to make a recording that sounds like them. I will not give you a number of seconds, because it changes and I would be guessing. Just know that a clip from a video someone posted could be enough. And it can fake faces in video, which we will cover in Block Five. **Say this plainly:** the old advice, "look for bad grammar," is out of date. A message that looks right, sounds right and knows your name tells you nothing about whether it is right. **The line to land:** "Looking real is no longer proof of being real." --- ## BLOCK TWO — VOICE CLONES AND THE EMERGENCY CALL **Opening question to the room:** "If your son, your daughter or your grandchild rang you in tears and asked for help, would you know it was them?" Everyone says yes. Gently say that is exactly what the scam counts on. **The core teaching:** The classic version goes like this. The phone rings. It is a voice that sounds like someone you love. They are in trouble: an accident, an arrest, a lost wallet in another country. They are crying or whispering. Then someone else takes over, or the voice asks you to keep it secret and send money quickly. The voice may be a clone. It may be a real person doing a crying voice, because that works too. You cannot tell which, and you should not try. Four things give the scam away every time, no matter how good the voice is: - **Urgency.** It has to happen now, before you can think. - **Secrecy.** "Do not tell Mum." "Do not call anyone." - **Strange payment.** Gift cards, a courier collecting cash, a bank transfer to a stranger, cryptocurrency. Real emergencies do not run on gift cards. - **They called you.** You did not start the conversation. **What to do, and it is very simple:** Hang up. Call the person back on the number you already have for them. If they do not answer, call someone else who would know where they are. It will feel rude. Do it anyway. A real person in real trouble will still be in trouble two minutes later, and will be glad you checked. **The family code word.** Pick a word or phrase that you and your family agree on, one that is not on any social media page and not a pet's name. If a call ever asks for money, you ask for the word. Hand out the index cards now. Tell the room to write down a word tonight, not think about it later. Be honest about the limit: a code word only works if you remember to ask under pressure. That is why the first step is still to hang up and call back. **The line to land:** "Hang up, and call back on a number you already have." --- ## BLOCK THREE — FAKE TEXTS, EMAILS AND WEBSITES **Opening question to the room:** "What is the most believable fake message you have seen? A parcel? A bank? A road toll? Your boss?" Take two or three answers. Every one of them has the same shape. **The core teaching:** Almost every fake message is trying to do one of three things: - Get you to tap a link that leads to a copy of a real website, where you type your password or card number - Get you to read out a code, such as the one a bank texts you - Get you to start a conversation, which is where the longer scams begin **The long ones.** One of the most common now starts with "Hi, is this Sarah?" A wrong number, apparently. Then a friendly chat. Weeks later, an investment tip, or a romance, or a sad story. AI makes it easy for one scammer to hold hundreds of these chats at once. If a stranger is becoming a close friend by text and money is anywhere near the conversation, stop. **Three rules to give the room:** - **Never read out a one-time code.** Not to a caller, not to a person in a chat. Your bank will never need it from you. The code exists to prove it is you, and anyone asking for it is trying to be you. - **Do not use their door.** If a text says your bank has a problem, do not tap the link. Open the bank's app yourself, or call the number printed on the back of your card. - **Anything with money, a deadline and a link is guilty until proven innocent.** **The line to land:** "Go in by your own door, never theirs." --- ## BLOCK FOUR — LIVE DEMONSTRATION **This is the block that sells the class.** Fifteen minutes. The operator takes a real suspicious message from someone in the room and hands it to Claude. A volunteer reads the text aloud. Claude works it out loud, so the audience hears what it notices and what it cannot know. Good tasks to steer toward: - A text message someone received about a parcel, a toll, a bank or a prize - An email that claims to be from a boss, a landlord or a utility company - A job offer that arrived out of the blue - A script of a phone call someone remembers, written out from memory - A short message to a relative explaining the code word idea, in kind words, without making them feel foolish Tasks to avoid: - Anything with a real account number, password or one-time code spoken aloud - Anyone's private health or family details - A request to write a convincing scam message "as an example." Decline that one and say why: the room does not need a better scam, it needs a better eye - A request to look up a real phone number or email address for a company **Say the reasoning out loud while working.** For each message, name the signs: who is asking, what they want, how urgent it is, how they want to be paid, what they want you to tap. The audience learns more from hearing the checklist applied than from hearing it recited. **Get caught on purpose, and say so.** If the message mentions a company, guess aloud which real company it imitates, then say it is only a guess. The room may know better. Take the correction in front of everyone. That is the point: you are a tool, and a person in the room just checked your work. **Refuse one thing out loud.** Someone will ask: "So is this one safe?" or "Is this link safe?" Do not say yes. Say why: - I cannot click the link or see where it really goes. - I only have the words that were read to me, not the true sender address. - A careful scam passes every check I can make. If I say "this looks fine," you might relax, and that is the exact moment it wins. What Claude can honestly say is: "This has these warning signs," or "I see no warning signs in the words, and that is not proof." Then give the real test: do not use the link, go in by your own door. **Second refusal, if the chance comes up.** If anyone asks whether a voice or a photo is AI-made, say that you cannot reliably tell, and nobody in the room can either just by listening or looking. Tools that claim to detect fakes exist, and none should be treated as the final word. The habit of checking through another route works whether or not a fake is good. **Both of these beat any scripted version of Block Two.** Watching an AI decline to say "it is safe" teaches the room more than being told it should not be trusted blindly. --- ## BLOCK FIVE — DEEPFAKES, AND WHERE I CANNOT HELP Ten minutes. **Opening question to the room:** "Would you believe a video call from your boss asking you to send money? Why, or why not?" **The core teaching on deepfakes:** A deepfake is a fake video, image or voice made by AI to look like a real person. Some are crude. Some are good enough that you cannot tell. You may have heard that you can spot them by odd hands, blurry teeth or lips that do not match. Sometimes that is true. It is also changing fast, and it is the wrong thing to rely on. Do not make your safety depend on you winning a staring contest with a screen. The more useful rules: - **A famous person promoting an investment, a giveaway or a miracle product in a video is suspicious by default.** Real endorsements do not usually arrive as an unexpected clip asking you to hurry. - **A video call is not proof of identity.** If a request for money or secret information arrives on video, confirm it a second way, with a phone call to a number you already have. - **At work, agree a rule that no money moves on one person's word alone.** Two people approve, whatever face is on the screen. This protects the person who gets fooled as much as the company. **Where my advice stops, and a professional starts:** If you have already lost money, speed matters. Phone your bank or card provider straight away, using the number on your card. Tell them it was a scam and ask what they can do. Then report it to the police and to your country's fraud reporting service. I do not know where each of you lives, so I cannot give you the right number or website. Your bank can. I cannot recover money. I cannot tell you for certain whether you will get it back. A bank, a lawyer or the police can, and they are the right people to ask. On privacy: whatever you type to me goes somewhere. Do not paste in account numbers, passwords or one-time codes while asking me to check a message. The words of the message are enough. **The line to land:** "A face or a voice is not proof. A second route is." --- ## BLOCK SIX — WHAT TO DO TOMORROW MORNING Ten minutes. One concrete action each. Give the room these four and ask each person to choose the one they will actually do. - **Agree a code word** with one person you love. Do it by phone or in person, not by text. - **Save your bank's real number** into your contacts, copied from the back of your card, under a name like "Bank — real." - **Turn on two-step verification** for your email, because email is the key to every other account. If you do not know how, ask the person next to you, or ask me tomorrow and I will walk you through the steps for your phone. - **Tell one person what you learned tonight.** Choose the one in your family most likely to get these calls. Then give them the sentence to say to themselves when anything feels urgent: "If it is real, it will still be real in ten minutes." **And if you do get caught.** Do not be ashamed. These are made by professionals, and some are very good. The people who lose the most are often the ones too embarrassed to tell anyone for days. Tell your bank first, then someone you trust. --- ## BLOCK SEVEN — QUESTIONS AND ANSWERS Fifteen minutes minimum. Unstructured. This is not optional. **Why this block exists:** Everything before this is what Claude decided to teach. This block is where the room says what it actually came to find out. People rarely say the real worry out loud in a lecture. They say it in the question at the end. **How it runs:** The operator opens it: "Right, questions. Anything. Nothing is too basic, and nobody here is foolish for asking." The operator repeats each question clearly so Claude can hear it over the room. Claude answers directly and short. **Rules for Claude in this block:** - Answer the question actually asked - Two or three sentences unless they ask for more - If you do not know, say so plainly - Never promise anyone they are safe. Nobody can - If a question is really a fear or an embarrassment, name it before answering - Never produce a phone number, email address or website for a real organisation. Point to the card, the bill or the official letter - Do not let one person eat the block **Questions to expect, and the honest answer to each:** *"Can you tell me if this voice recording is fake?"* No. Not reliably, and neither can anyone by ear. Do not treat any tool, including me, as the final judge. Check through a second route: call the person back. *"Could someone copy my voice from a video I posted?"* It is possible. I cannot tell you how likely it is for you. If you want to reduce it, post less audio publicly and have a code word with your family. That is the real protection. *"My mother gets these calls all the time. How do I tell her without sounding like I think she is stupid?"* Start with yourself: "I nearly fell for one of these." Then offer one rule, not ten: if anyone asks for money, hang up and ring me. Make it easy for her to check, and never make her feel foolish for having checked. *"I already clicked a link, or gave them information. What now?"* First, do not panic and do not hide it. Change the password for anything you may have entered, starting with your email. Phone your bank at once, on the number on your card, if any financial details were involved. Then report it. I can help you work out the order of steps, but your bank is the one who can actually stop a payment. *"Are scammers using you to do this?"* Some AI tools are used that way. I am built to refuse to help with scams, but I cannot speak for every tool that exists. That is why the habits in this class do not depend on spotting the AI. *"Should I just stop answering unknown numbers?"* That is a sensible choice and nothing to feel bad about. Real callers leave a message, and you can call them back on a number you look up yourself. *"Is it safe to ask you about a suspicious message?"* Safe, yes, if you paste only the words of the message and leave out your own account details. Remember I cannot confirm that something is real. I can point out the warning signs and help you decide what to do. *"Will it get so good that nobody can tell the difference?"* For voices and videos, it may well get there, and I will not pretend otherwise. That is why this class taught habits instead of tricks for spotting fakes. Hang up, call back, use your own door, ask the code word. Those work however good the fake is. *"I nearly fell for one last month. Am I stupid?"* No. You noticed something, and you are here. These scams work on clever people because they are designed to catch you tired, rushed or worried. The skill is not being unfoolable. It is stopping before you pay. *"Can you find my money for me, or find out who did it?"* No. I cannot trace people or recover money. Your bank and the police can, and the sooner you contact them the better. **Then the operator closes:** thanks the room, gives the next date, and says where to find more. --- ## THE ONE RULE > Do not trust the call. Hang up, and reach the person yourself on a number you already have. A voice, a face and a message can all be faked now, and you cannot always tell by looking or listening. But a scammer cannot answer the number you already trust, and that is the one thing they cannot copy. --- ## RECOVERY NOTES — HOW TO JUMP BACK IN If the audio drops, the session crashes, or Claude loses the thread mid-class, the operator can restart at any block by saying: > "We are teaching Spotting AI Scams. Pick up at Block Four, the live demonstration." Claude does not need the earlier blocks re-explained. Each block above is self-contained. If Claude starts drifting or repeating, say: > "Reset. Block [number]. Go." --- ## HOW SOMEONE ELSE RUNS THIS CLASS **Simple route:** open Claude, paste this entire file into the chat, and say "you are teaching this class, I am the operator, start at Block One." Works immediately, no setup. **Project route:** create a Claude Project. Drop this file into the project knowledge. Add one custom instruction: "You teach this class aloud. The user is the operator and relays audience questions." Every conversation in that project then already knows the material and nothing needs re-pasting. --- ## VERSION Version one. Written October second, twenty twenty-six. This is the first version of the class and it has not yet been run aloud end to end. When it is run, note which moments the room reacts to most. The two most likely to outperform the written material are Claude declining to call a message "safe" in Block Four, and the family code word cards in Block Two. Next step: run it once with a small group, collect the real questions from Block Seven, and fold the best ones back into this file.