Deepfakes and phishing: artificial intelligence in the service of digital fraud
74% of workers in the EU have encountered suspicious messages, whilst only 48% believe they can identify AI-manipulated video.
Artificial intelligence and cybercrime: figures that raise concern
74% of employees in the European Union have encountered suspicious emails, messages or links in the workplace, according to a Eurobarometer survey published by the European Commission on 30 September 2026. Only 48% believe they can identify video manipulated with artificial intelligence (AI), whilst 15% reported specific incidents of fraud involving AI-generated content. These figures are published and discussed by In.gr, amid the European Cybersecurity Month in October.
Phishing enhanced by generative AI
Phishing has long been one of the dominant methods of cybercrime. According to the ENISA Threat Landscape 2025 report, it accounted for approximately 60% of the identified initial methods of system access examined. Europol warns that generative AI tools now allow the drafting of personalised messages, with the use of personal information and adaptation to the linguistic and cultural context of each potential victim.
An SMS can be presented as a package delivery notification, whilst an email can request confirmation of banking details with a professional appearance and absence of spelling errors. These characteristics are no longer reliable indicators of authenticity.
Deepfakes: voice and video cloning
Deception becomes even more complex through deepfakes — synthetic or manipulated videos and audio clips that present faces saying or doing things that never happened. Voice cloning technology can reproduce the voice of a relative or colleague, creating a false sense of trust. Manipulated videos are used to promote false investments or to impersonate public figures.
The threat reaches the user through video calls, social networks or everyday applications. The most dangerous element of these attacks is not their technological complexity, but their ability to provoke fear, trust or a sense of urgency.
E-commerce and investment fraud
The use of AI is also expanding into e-commerce. Fake online stores, misleading advertisements and websites promising quick profits combine with well-designed content, creating an apparently reliable digital presence. According to the ENISA Threat Landscape 2026 report, published in September, online investment fraud is estimated to have cost approximately 4 billion euros in the European Economic Area in 2024, according to figures cited by the organisation from the European Banking Authority.
An attractive offer is not in itself an indication of fraud, but when accompanied by pressure to purchase immediately, further verification is required. Attention should be paid to the identity of the seller, the actual address of the website and unusual payment requests.
Protection measures: from technical security to verification
Protection against new threats is not limited to installing security software. Regular device updates, strong and unique passwords, activation of multi-factor authentication and reliable antivirus programmes are fundamental measures. Also useful is the updating of router software, as well as maintaining backup copies on external SSD drives or appropriately protected USB sticks — a measure that mainly contributes to data recovery following technical incidents.
The most important defence, however, remains independent verification. A suspicious request should be verified through an official communication channel, without using the links or telephone numbers contained in the message itself. In the event of disclosure of banking details, immediate contact with the bank is required.
As artificial intelligence makes digital deception easier to produce and harder to identify, trust can no longer be based solely on what we see or hear. Verification before any financial transaction or disclosure of personal data emerges as a fundamental prerequisite of digital security.
Source: In.gr