Safety & Fraud Protection
What to do when the voice on the phone sounds like a relative but isn't one.
The phone rings. It is your granddaughter's voice — she is crying, she needs money urgently after an accident. It sounds exactly like her. But it is not her — it is a computer. AI can now clone voices, write perfect phishing emails, and even generate fake video calls.
The old tricks for spotting scams (bad grammar, strange voices) no longer work. But the good news: five simple rules protect you — no technical knowledge required.
Grandchild Scam 2.0: When AI Clones Voices
The classic grandchild scam has been known for decades: a stranger calls and pretends to be a relative. In the past, you could often detect the fraud by the wrong voice. That has fundamentally changed.
Definition
Voice Cloning
Analogy
Definition
A stranger's voice. Hopes you don't notice the difference. Often transparent.
Cloned voice of your real grandchild. Sounds exactly like the real person. Hard to tell apart for the human ear.
A typical case: A grandmother receives a call from "her grandson." He is crying, says he had a car accident and needs 8,000 euros for bail. The voice was generated from a 10-second clip from his Instagram. She protected herself by hanging up and calling her grandson at his saved number — he was safely at work.
Why your ears deceive you: AI does not just copy the pitch — it replicates micro-patterns in speech (breathing, hesitation, intonation) that your brain uses to identify people.
Caution
Misconception: I Would Recognize a Fake Voice
To the human ear, modern voice clones are hard to tell apart from the real voice. Your behavior (callback) protects you, not your hearing. The AI reproduces exactly the fine patterns your brain uses for voice recognition.
Just 3 to 15 seconds of audio can be enough — for example from a social media video, a voice message, or a public talk. Services like ElevenLabs create an AI model of the voice from this; with one to two minutes of audio, the copy gets more accurate.
The criminal then types text or speaks — and the system outputs everything in the cloned voice. In real time. The quality is so high that even close family members often cannot hear the difference.
AI Phishing: Perfect Forgeries by Email
Besides phone calls, scammers also use email and SMS. In the past, you could easily spot phishing messages by their spelling mistakes and generic greetings. AI has fundamentally changed that.
Definition
Phishing
Analogy
Definition
Bad grammar, spelling mistakes, generic greeting ("Dear Customer"). Often recognizable at first glance.
Perfect grammar, your real name, your real bank, professional layout. Virtually indistinguishable from genuine bank communication.
A phishing email reads: "Dear Mr. Schmidt, we detected a suspicious transaction of 499.00 euros on your account. Please confirm your identity at the following link within 48 hours." Perfect language, personalized name, realistic amount.
The three red flags: (1) Artificial urgency ("48 hours"), (2) you are asked to click a link, (3) the sender's or the link's domain is wrong (sparkasse-sicherheit-check.ru instead of sparkasse.de). Your bank never asks for passwords or TANs by email.
Caution
Misconception: I Spot Phishing by Grammar Mistakes
With AI, grammar errors can be avoided completely. The BSI explicitly warns: checking for spelling mistakes is no longer enough to spot phishing. Instead, check: the sender address, the domain of any link, and whether the email creates artificial urgency.
According to the FBI, people aged 60 and over reported more internet crime and higher losses in 2025 than any other age group: about 7.7 billion US dollars. Millions of phishing emails are sent worldwide every day.
The 2024 situation report of Germany's Federal Office for Information Security (BSI) states that attackers use AI language models to write the text of phishing messages. In a survey by the BSI and the TÜV association (October 2026), 90 percent of companies with AI-related attacks or suspected attacks named deceptively genuine phishing emails.
Interactive: Is This a Scam Attempt?
You now know the methods scammers use: cloned voices, perfect phishing emails, and fake video calls (so-called deepfakes). Click through the decision tree to identify the type of fraud and learn how to protect yourself.
Five Golden Rules: Your Shield
You don't need to understand the technology to protect yourself. These five behavioral rules work regardless of how convincing the forgery is — because they target your actions, not the attacker's technology.
Think of these rules like locking your front door before leaving the house. You don't need to understand how a burglar picks a lock — you just need to always turn the key. These rules work the same way: follow them consistently, and the scam cannot reach you.
Audio for Voice Clone
3 -15 seconds
Can be enough for a deceptively similar copy of your voice
Phishing Emails
Millions daily
Sent worldwide every day
Primary Target Group
60 + years
Seniors are disproportionately targeted
Note
Five Golden Rules
- 1. Callback Rule: Hang up and call back at the saved number — never the displayed one.
- 2. No-Rush Rule: Legitimate institutions never pressure you for immediate action.
- 3. No-Data-by-Phone Rule: Never share passwords, PINs, or TANs on the phone.
- 4. Four-Eyes Rule: Always discuss transfers and cash handovers with a trusted person first.
- 5. Link Rule: Never click links in emails or SMS from banks — manually type the URL in your browser.
Note
How to Apply the Rules
Your phone rings. A voice that sounds exactly like your daughter says she had an accident and urgently needs 3,000 euros.
You say: "I'll call you right back" and hang up (Rule 1). You call your daughter at her saved number. She answers — she is fine. Scam neutralized.
If she does not answer: call another family member (Rule 4). Under no circumstances transfer money while in a state of panic (Rule 2). Prevention helps: agree on a secret code word with your family. Anyone on the phone who cannot say it is suspicious.
Caution
Misconception: The Police Can Get My Money Back
Money transferred to foreign accounts or converted to cryptocurrency is almost always gone permanently. Prevention is the only reliable protection. If it has happened anyway: call your bank immediately and report it to the police. Only if you act fast can a transfer still be stopped, and even then only in exceptional cases.
These rules will continue to work in the future. No matter how convincing the technology becomes — the rules target your behavior, not the technology. They are future-proof.
Interactive: Can You Spot the Scam?
You now know the threats and the five golden rules. Test your knowledge with four realistic scenarios — which rule do you apply?
Interactive: How an AI Detects Fraud
Banks check millions of payments every day. No human can do that — but an AI can. It looks for whatever falls out of the ordinary. In the field below, each dot is a payment: most sit close together (normal), a few are lonely outliers. Try out how the AI spots the outliers — and where the line between suspicious and harmless runs.
Note
Takeaways
- AI can clone any voice from a few seconds of audio — your ear alone can no longer reliably tell real from fake.
- AI-generated phishing emails have perfect grammar, use your real name, and look professional — old detection tricks (spelling errors, generic greetings) no longer work.
- Five simple behavioral rules protect you against both threats — no technical knowledge required.
Check
Quiz: Safety & Fraud Protection
What makes the AI grandchild scam more dangerous than the classic version?
Select one answer
1. What makes the AI grandchild scam more dangerous than the classic version?
2. You receive a perfectly written email from "your bank" asking you to confirm your identity via a link within 24 hours. Which three signs suggest phishing?
3. Your phone rings. A voice that sounds exactly like your son says he had an accident and urgently needs 5,000 euros. What do you do first?
4. A neighbor says: "I always check for grammar mistakes — that's how I know an email is fake." Why is this strategy no longer reliable?
Check
Checkpoint: Safety & Fraud Protection
Can you explain this in your own words? Tick what you've got.
0 of 3 done