Treat every unexpected call, message or video that asks for money, a code or quick action as a possible deepfake, and check it through a second channel before you do anything. AI can now copy a voice from a few seconds of audio, write a flawless message in perfect Dutch or French and put a familiar face in a video, so the old warning signs like bad spelling or a strange accent no longer protect you. This guide lists the clues that still work for calls, messages and videos, and the one check a scammer cannot control: hanging up and calling the person back on the number you already have saved.
What is a deepfake?
A deepfake is a video, image or audio recording that has been created or changed with artificial intelligence to make someone appear to say or do something they never said or did. The word combines "deep learning", the AI technique behind it, and "fake". Deepfakes started as face swaps in videos, but today the term also covers cloned voices on the phone and AI written messages that imitate a real person.
How to tell if a call is AI generated
Hang up on any call that pushes you to pay, share a code or confirm something in your banking app, even when the voice sounds exactly like someone you know. A scam call with a cloned or AI voice is a form of vishing, short for voice phishing, and it is the deepfake you are most likely to meet on your gsm. Listen for these signs:
Pressure and urgency. The caller needs money, a payment confirmation in your banking app or a code from an SMS right now, and asks you not to tell anyone.
Slightly flat or robotic speech. The voice may lack natural breathing, emotion or the small sounds people make while thinking.
Odd pauses. Live AI voices often wait a moment before each answer while the system generates the reply.
Too little or too perfect background noise. A real call from a busy street rarely sounds like a quiet studio.
Vague answers to personal questions. A clone can copy a voice, but it does not know what you had for dinner last Sunday.
Remember that the number on your screen proves nothing. Caller ID can be spoofed, so a call can show a familiar number or even your bank's real number while coming from someone else entirely.
How to tell if a message is AI generated
Judge a message by what it asks you to do, not by how well it is written, because AI now writes flawless text in any language and good grammar no longer proves anything. Watch for these patterns:
A new number with a familiar story. The classic is "Hi Mum, my phone broke, this is my new number", followed by a request to pay an urgent bill.
A link or an attachment you did not expect, often to "confirm", "unlock" or "update" something.
A refusal to call. The sender always has a reason why a phone or video call is impossible right now.
Emotional pressure. Fear, guilt or excitement are used to stop you from thinking twice.
How to tell if a video is AI generated
Slow the video down and look at the eyes, the lips and the edges of the face, because that is where a deepfake video still shows its mistakes, whether it features a celebrity promoting an investment or a family member on a video call. Check these details:
Eyes and blinking that look unnatural, or a face that hardly moves while the person talks.
Lips that do not match the words.
Blurry edges around the face, hair, ears or glasses, especially when the head turns.
Lighting and shadows that do not match between the face and the background.
Hands, teeth and jewellery that change shape or disappear.
Then check the source. Search for the same clip on trusted news sites, run a reverse image search on a screenshot and look at who posted it first. A video with a famous face that only exists on one unknown account is a strong warning sign.
Deepfake warning signs at a glance
Format | Common warning signs | Best check |
|---|---|---|
Call | Urgency, flat voice, odd pauses, a familiar number | Hang up and call back on the number you have saved |
Message | New number, unexpected link, no calls possible | Call the person on their old number or ask a personal question |
Video | Unnatural blinking, lips out of sync, blurry edges | Look for the same video on trusted news sites |
The check that always works
Verify every unusual request through a second channel before you act. End the call or chat, then contact the person or company yourself on a number or app you already trust. This works because it does not depend on spotting anything at all, while warning signs and detection tools fall behind every time the technology improves. A real family member will understand, and a real bank will never ask you for codes or payment confirmations over the phone.
Many families now agree on a safe word, a short word or question that only they know and that anyone can ask during an unexpected call. It takes a minute to set up and works against even the most convincing deepfake voice.
What to do if you think you were targeted in Belgium
Stop the conversation, do not pay and do not share any code, then use the right contact for your situation:
Situation | What to do |
|---|---|
You shared bank details or transferred money | Call your bank or Card Stop on 078 170 170 right away |
You received a suspicious email | Forward it to suspicious@safeonweb.be, run by the Centre for Cybersecurity Belgium |
You lost money or were threatened | Report it to the local police and keep screenshots, numbers and messages as evidence |
Telecom that works for you
Verify first and act second, and expect the same straight answers from the company behind your gsm. Firsty is building telecom that works for you, not against you: no SIM card to swap when you travel, your own number that stays reachable wherever you are and no surprise bill. Every connection Firsty sells also connects someone in Belgium who needs it, not as charity, but as how the business works. Things are moving, so keep an eye on what Firsty is building and discover more at Firsty.app.
FAQ
What is a deepfake?
A deepfake is a video, image or audio recording that has been made or changed with artificial intelligence so that a real person appears to say or do something they never did. The term comes from "deep learning" and "fake", and today it also covers cloned voices used in scam calls.
Can AI clone a voice from a short clip?
Yes. Current voice cloning tools can copy the sound of a voice from just a few seconds of audio, for example from a video posted on social media or a voicemail greeting. A cloned voice can sound convincing in a short call, which is why verifying through a second channel matters more than trusting the voice.
What is vishing?
Vishing, short for voice phishing, is a scam where a criminal calls you and pretends to be someone you trust, such as your bank, the police or a family member, to get money, codes or personal data. AI makes vishing more convincing because scammers can now use cloned voices and spoofed caller ID.
Is there an app that detects AI generated calls or videos?
Some AI voice detector and deepfake detection tools exist, but none of them is reliable enough to trust on its own, and new AI models regularly fool them. Use them as an extra signal at most, and always verify an unusual request by contacting the person or company yourself.
Do AI generated videos have to be labelled in the EU?
Yes, under Article 50 of the EU AI Act, organisations that publish deepfakes must disclose that the content is artificially generated or manipulated, and this obligation applies from 2 August 2026. Scammers ignore these rules, so a missing label never proves that a video is real.
Where do I report a deepfake scam in Belgium?
Forward suspicious emails to suspicious@safeonweb.be and report fraud to your local police, especially if you lost money. If you shared bank details or made a payment, call your bank or Card Stop on 078 170 170 immediately to block your cards.
