Artificial Intelligence (AI), like a coin, has two sides. While it profoundly transforms the way we communicate, work, and live, becoming more efficient and innovative, it also serves as a highly useful resource for cybercriminals, who have successfully created schemes that are increasingly difficult to detect.
Cloned voices, highly personalized automated messages, and fake videos that appear real are just a few examples of scams that leverage AI to manipulate emotions and exploit victims' trust. The goal is almost always the same: trick them into sharing confidential information, making financial transfers, or clicking on malicious links in order to steal personal data, for example.
Among the most common types of fraud, already recognized and identified, are forgeries, in English deepfakes, you voice cloning attacks, with examples even of voices close to the target person, you scams through relationships online and the fake business emailsThese threats use data that is often publicly available to create compelling content and use the urgency to make it difficult to discern and distinguish between what is real and what is fake.
To protect yourself, always maintain a vigilant posture:
- Stop and check, before acting on urgent or unusual requests.
- Confirm identities, through direct contacts and official channels.
- Limit the sharing of personal data on social media.
- Use safe words between family members or work teams.
- Report suspicious situations to cyber support services or to competent authorities.
This article is part of the GÉANT's cybersecurity campaign “Be mindful. Stay safe””, which aims to raise awareness among users and organizations about the risks associated with AI-driven digital fraud. FCCN, FCT's digital services unit, participates in this dissemination partnership, reinforcing its commitment to the digital security of the academic and scientific community in Portugal.
In a world where digital manipulation becomes increasingly convincing, attention and verification remain the best protection tools.
