Seeing and hearing is no longer believing
In short: a deepfake is a video, a voice or an image created or altered with artificial intelligence to make it look as if a real person said or did something they never said or did. Scammers use them to convince you to invest, to send money to a “relative” in trouble, or to make an urgent payment at work. Do not rely on what you see and hear: check through an official channel, call the person yourself on a number you know, and do not act under pressure.
Table of Contents
What is a deepfake
The word comes from “deep learning” (a branch of artificial intelligence) and “fake”. An AI program analyzes hundreds or thousands of images and recordings of a person, then generates new video or audio in which that person appears to say whatever the creator wants.
Early deepfakes were easy to spot: blurry faces, lips that did not match the voice. Today’s are far more convincing, and a few seconds of audio, taken from social media or a voice message, are enough to clone a voice. The tools are cheap and easy to use, so you no longer need to be an expert to make one.
How scammers use deepfakes
- Ads with famous people “recommending” investments. Politicians, business people, TV presenters or doctors appear in altered videos praising an investment platform or a “cure”. In Romania, for example, the national cybersecurity authority warned about deepfake videos of the country’s president and prime minister promoting “investment opportunities”. Read more in trading platform scams.
- The voice of a relative in trouble. You get a call or a voice message with the voice of your child, parent or a friend: “I had an accident, I urgently need money”. The voice is cloned from public recordings.
- The boss who asks for an urgent transfer. At work, you receive an email and then a video call with your “manager” or “CFO” asking for a confidential payment. In 2024, an employee of the engineering firm Arup in Hong Kong transferred about 25 million dollars after a video call in which every “colleague” was a deepfake.
- Fake profiles and blackmail. Generated photos and videos for fake dating profiles, or faked intimate images later used for blackmail.
- Disinformation. False statements put in politicians’ mouths, especially during election campaigns, to manipulate public opinion.
What's new in 2026
- Deepfakes must be labelled in the EU (August 2026). Since 2 August 2026, the transparency rules of the EU Artificial Intelligence Act (AI Act) apply: anyone who publishes a deepfake must clearly disclose that the image, voice or video was generated or altered with AI. Some technical marking obligations for systems already on the market were postponed to 2 December 2026. Scammers will not follow the rule, but a missing label can no longer be blamed on ignorance.
- National warnings about deepfake content (August 2026). On 26 August 2026, Romania’s National Cyber Security Directorate (DNSC) warned about AI-generated or altered images, voices and videos used for disinformation, manipulation and fraud. Its advice: check the source, look for unusual details in the image or sound, consult several independent sources and do not share shocking content before confirming it is real.
- A well known doctor used in ads without consent (September 2026). A Romanian investigation found that ads for “miracle” treatments shown on news websites used AI-generated videos of a well known surgeon. Read more in medical scams.
The practical lesson: in 2026, a video of a famous person is no longer proof. It is just a claim that needs checking.
How to recognize a deepfake
Good deepfakes are hard to spot with the naked eye, but many still show signs:
- The lips do not quite match the voice, especially on fast words.
- The face is too stiff or too smooth, blinking is rare or unnatural, the gaze looks “empty”.
- Odd edges around the face, hair, glasses or earrings, especially when the person turns their head.
- The voice sounds flat or metallic, with no breathing and an intonation that does not fit the emotion of the message.
- The background or lighting does not match the face.
The surest sign is not technical, it is the context:
- The message asks you for money, data or a quick decision.
- The person does something they normally would not: a president does not recommend investment platforms, a doctor does not sell cures in ads, a CFO does not ask for secret transfers on WhatsApp.
- The video does not appear on the person’s official channels or in reputable media.
How to protect yourself
- Agree on a family safe word. A word only you know, which you ask for in any urgent call about money.
- Hang up and call back on the number you already have saved, not the one that called you.
- Check official channels. Look for the statement on the official website or pages of the person or organization.
- At work, use the two-confirmation rule for any urgent or unusual payment, through a second channel, no matter who seems to be asking.
- Do not rush. Urgency is the scammers’ main tool; a few minutes of checking will not harm a real situation.
- Limit what you publish. Public video and voice recordings can be used for cloning; make your profiles private where you can. See also how to protect your social media accounts.
What to do if you were tricked, or if you appear in a deepfake
- If you sent money, call your bank immediately to block or dispute the payment, then file a police complaint.
- Keep the evidence: the video, the link, screenshots, the phone number and the conversations.
- Report the content to the platform where you saw it and tell your national cybersecurity authority (in Romania, DNSC at 1911 or pnrisc.dnsc.ro).
- If you appear in a deepfake, ask the platform to remove it and warn your contacts. For faked intimate images, StopNCII.org helps you stop them spreading on major platforms without sending the images to anyone.
Useful tools
No tool is infallible, but these can help with a first check:
- InVID-WeVerify: a browser extension journalists use to analyze videos and trace where they come from.
- Google Lens and TinEye: reverse image search, to find the original of a photo or a video frame.
- Deepware Scanner: scans a video for signs of a deepfake.
- Sensity: a deepfake detection platform, used mostly by companies.
And the most important tool is still a simple question: “Who gains if I believe this right now?”. The same pattern appears in phishing.
Frequently asked questions
What is a deepfake, in simple terms?
A video, a voice or an image created or altered with artificial intelligence to make it look as if a real person said or did something they never said or did.
Can a video of a famous person recommending an investment be real?
Almost certainly not. Politicians, bankers and celebrities do not recommend investment platforms in social media ads. Check their official channels.
How do I protect myself from a call with a relative’s cloned voice?
Hang up and call the person yourself on the number you know. Agree on a family safe word for emergencies in advance.
Where do I report a deepfake?
To the platform where you saw it, to your national cybersecurity authority, and to the police if you lost money.
Sources
- Paul, Weiss: EU Finalises Transparency Rules for AI-Generated Content (August 2026)
- European Commission: Code of Practice on Transparency of AI-generated Content
- CNN: Arup revealed as victim of $25 million deepfake scam (May 2024)
- StopNCII.org: Stop non-consensual intimate image abuse
About the author
Brîndușa Maria, Senior Software Test Analyst, ISTQB® Certified Tester Advanced Level Test Analyst (CTAL-TA). I write about how to recognize online scams. LinkedIn
Note: this article is based on my research and professional experience. I used artificial intelligence tools for drafting and editing.
📢 Share the knowledge, stay safe! 🔒✨ Use the share buttons below and help others stay safe online. The more we know about online scams, the better protected we are! 🚀💙
