Associated Incidents
AI-generated videos impersonating President Nicos Christodoulides, business tycoons and celebrities have scammed Cypriots out of thousands of euros, with authorities warning citizens have no protection against the growing threat.
Around 15 people in Cyprus lost between €10,000 and €15,000 after falling for a deepfake video that showed Christodoulides alongside politicians Averof Neophytou and Haris Georgiades promoting a fake investment platform promising prosperity for all.
Giorgos Michailidis, Commissioner of the Digital Security Authority, said no tools or mechanisms exist to protect people from malicious AI.
"It's very difficult for someone to protect themselves. They need to be very observant. Look at the hands, the fingers, for example. It's already very difficult, and I believe in the coming years it will become even more difficult as artificial intelligence develops and gets better and better," Michailidis said.
Scammers have deployed high-profile figures to lend credibility to their schemes across several months.
In March, Sir Stelios Haji-Ioannou, founder of easyGroup, warned the public about a cryptocurrency scam using his image and company name.
The misleading online advert urged people to register with a fake company called "easyProfit", offering returns of up to 96 per cent from cryptocurrencies and foreign exchange trading.
Singer Anna Vissi and European Central Bank President Christine Lagarde also appeared in similar investment scheme videos.
In July, businessman John Christodoulou posted a scam video on his personal account showing him urging people to participate in frauds. "It's not me," he clarified.
In August, a fake Facebook page impersonating Phileleftheros used Greek presenter Sakis Tanimanidis to urge citizens to download an app promising large financial returns.
At Easter, a video circulated showing Donald Trump wishing Cypriots "good souvla" and promising to drag the Turks into the sea, all spoken in English. MP Christos Christofidis said countless people sent him the video, enthusiastic that Trump would save Cyprus.
"You can't imagine how many people sent it to me excited that Trump would come save us," Christofidis said.
AI scams can take many forms, Michailidis said. They can be voices in phone calls, videos, photos, written text in chats or other messages. Someone communicating on Instagram or Facebook might be talking to AI rather than a real person.
Even experts lack tools or mechanisms to detect them, he said.
During recent major earthquakes in Turkey, a photo circulated of a Greek rescuer holding a child he had supposedly just saved. But anyone looking at the rescuer's hands would notice he had six fingers, Michailidis said.
No manual exists for what exactly people should look for, he said. They need to watch for anything that doesn't look right.
The European Union has opened discussions on mandatory labelling for AI-generated content, Michailidis said. But this only applies to legitimate AI tools used in good faith. Anyone using it maliciously to deceive people won't comply with labelling requirements.
Citizens must verify information, he said. "It would be good for people to call back to make sure."
AI technologies are advanced, Michailidis said. Someone can pay a subscription to acquire malware and viruses to attack a person or company. "It's unbelievable how much criminality in the AI sector has advanced," he said.
Social engineering through AI is happening more frequently, Michailidis said. These are actions through AI aimed at manipulating people, influencing them psychologically and directing them towards specific actions and behaviours. Something like this appears to have happened in the US presidential election that elected Donald Trump, as well as in the UK's Brexit referendum, he said. A state could be behind such actions, not just fraudsters.
The most important action now is informing citizens, Michailidis said. "People must not believe what they see, even if it looks real. They need to check it," he said.
"If we're not careful, we're at the mercy of artificial intelligence. We need to check and search. It depends on us," he said.
Christofidis recently submitted a law proposal criminalising the use of people in AI material without the affected person's consent.
"Recently I saw a video with a well-known presenter talking to people about a weight loss product. I watched it three times -- it was so believable -- to conclude it was AI-edited. If today we can tell, in one or two years when it's perfected, it will be impossible to realise," Christofidis said.
A gap exists in legislation to protect citizens, he said.
The European Union must require social media platforms to label AI material, Christofidis said. He expressed concern that propaganda videos and revenge videos will follow.