Deepfake alerts: how criminals use AI to commit fraud on social media

NordVPN has warned that videos, images and audio manipulated by artificial intelligence are being used in digital fraud. The company recommends that users verify the source of such content and confirm the identity of the person presenting it through other channels, rather than relying solely on visual flaws to detect a fake.

A familiar face and a familiar voice can make a fake offer seem trustworthy. According to NordVPN, scammers use deepfakes to promote fraudulent investments and fake products on social media, attributing recommendations to real people that they never made.

The technique, known as deepfake, involves creating or altering images, videos or audio using AI to mimic someone’s appearance, voice and behaviour. The digital security company believes that these forgeries increase the risks of identity theft, extortion and damage to reputation.

The scam exploits the trust placed in someone who appears to be speaking. Instead of presenting just a promise in text, the fake can put that message in the mouth of a professional or a personality recognised by the public.

How does a recording become a speech that never existed?

Real photographs, videos and recordings serve as the basis for the algorithms that create the fake. The systems learn the characteristics of the person depicted and reproduce them in new content.

“The process typically involves training algorithms with real images, videos or recordings of a person, allowing them to reproduce facial expressions, movements and vocal characteristics in a convincing way,” says Marijus Briedis, NordVPN’s chief technology officer, in a statement.

The technology also has legitimate applications in entertainment, audiovisual production and education. The risk pointed out by the company lies in its unauthorised use to deceive the public or harm the person whose image has been reproduced.

What signs can indicate a deepfake?

According to Briedis, unnatural facial movements, inconsistencies in lighting and blurred edges around the face can indicate manipulation. Changes in audio quality and differences compared to known recordings of the person also deserve attention.

These clues help raise suspicions, but they do not replace a complete verification. The executive stresses that identifying forgeries becomes more difficult as the technology evolves. The absence of a perceptible defect should not be treated as confirmation of authenticity.

To assess possible AI-based scams, NordVPN recommends verifying the source of the content and comparing it with reliable sources. Before believing an offer or sharing a statement, it is necessary to check if it appears on the official channels of the person who supposedly made it.

Measures to reduce exposure to fraud

The company’s recommendations include:

• Limiting the disclosure of personal information on the internet;
• Confirming the identity of interlocutors in suspicious communications;
• Activating multi-factor authentication on digital accounts.

“Digital education and privacy protection represent the main defences against attacks based on artificial intelligence,” says Briedis.

The measures serve different functions. Protecting an account makes it harder to access it improperly; verifying the source of a message helps assess who is behind it. A convincing appearance alone does not solve this second task.

Frequently Asked Questions

Are deepfakes just videos?

No. They can also be images and audio manipulated or produced with artificial intelligence to simulate people.

Is it possible to recognise all fakes by looking at them?

No. Flaws in the face, lighting or audio can raise suspicions, but verification should also consider the source of the content and reliable sources.

Does the alert measure the growth of deepfakes in Brazil?

NordVPN’s statement does not provide data that allows quantifying this growth in the country.