How scammers are exploiting artificial intelligence tools

Latvia's Safer Internet Centre says artificial intelligence has fundamentally changed the nature of online fraud — the poorly written scam letters of the past are largely gone. In early February, to mark Safer Internet Day, a conference in Latvia held the country's first Alan Turing test, in which 220 teachers and students tried to tell apart answers given by well-known public figures from those generated by AI. Participants got it wrong in nearly 41% of cases, well above the 30% threshold that, under Turing's own criteria, would mean the AI had effectively won.

What the technology can do

According to experts, just a few seconds of a publicly available audio recording is enough to create a near-identical voice clone that can be used in fraudulent calls. Tools such as ChatGPT, and unrestricted black-market versions of similar systems, allow criminals to produce grammatically flawless, personalized emails in any language. Even without deep technical skills, fraudsters can quickly build fake websites or malicious code, while automated analysis of people's social media profiles helps generate psychologically tailored messages aimed at specific targets.

Common tactics include deepfake videos of recognizable public figures supposedly promoting investments or cryptocurrency schemes, urgent calls using a cloned voice to request money transfers, highly personalized phishing emails, fake news websites publishing misleading articles, and armies of AI-generated social media profiles designed to create a false sense of popularity and trust.

How to protect yourself

Karlīna Šadre of the Drossinternets.lv education programme recommends several precautions. Families can agree on a secret word or phrase to confirm identity if a suspicious call claims to come from a relative in distress. Flawless language is no longer proof of authenticity — attention should shift to content rather than style, especially if a message conveys urgency or promises unrealistic profits. In video calls, telltale signs of manipulation can still include unnatural blinking or lips not matching the audio; asking the other person to turn their head or cover part of their face can expose an AI-generated image. Two-factor authentication remains effective, and requests for authorization codes should never be approved unless the person initiated the action themselves. Finally, people should avoid making rushed decisions when pressured to transfer money or share data.

Experts conclude that artificial intelligence itself is not inherently dangerous — the risk lies in trusting content without verifying its source. Critical thinking, they emphasize, remains the most effective defense against such fraud.