Models Looking for Work as an "AI Face Model" Become Tools for Scammers

Thousands of people look for side income online, and some stumble on an unusual offer: become an "AI face model" and earn money simply by joining video calls, 100 sessions a day. The work looks strange but harmless. In reality, it sits on top of an industrial-scale fraud industry.

Journalists at WIRED examined dozens of Telegram channels and found systematic ads recruiting "AI models": people whose faces will be used in schemes involving artificial intelligence. The channels vary in character. Some present themselves as freelance platforms, others as video-model agencies. Most candidates are women. Employers promise a flexible schedule, high pay and simple requirements: a good appearance, comfort on camera and willingness to work from morning to night. The real purpose of the job is never mentioned.

According to the investigation, the people who get hired unknowingly become the "face" of classic fraud schemes: romance scams, investment fraud and financial manipulation. Their images and videos are used to create deepfakes or as a live front: the victim sees a real person on screen and lets their guard down. As a result, people send money to someone who never existed.

The scheme works in several ways. In some cases the "model" runs the call herself from a pre-written script, effectively acting as the scammer even though she does not fully realize it. In others, her image or video stream is used in real time through AI face-swap tools, while someone off camera speaks the actual words. The second variant is technically convenient because the "model" formally "did not deceive" anyone, even though in practice she took part in a crime and supplied her biometric data for illegal use.

The technical barrier for such operations has dropped sharply over the past two years. Real-time face-swap tools, including DeepFaceLive, various forks built on InsightFace and other solutions, are free and run on an ordinary gaming PC. Face-swap quality has improved to the point where victims struggle to spot the trick, especially on an unstable connection, which scammers can deliberately degrade. Basic setup takes a few hours and requires no special knowledge. Launching such a scheme no longer takes a professional studio: a Telegram channel and a few hundred dollars to pay performers are enough.

It is important to grasp the scale. Telegram channels with "AI face model" job listings have thousands of subscribers, and some operate in the open. The ads openly mention "emotional communication with clients," "building trusting relationships online" and high pay for "effective" calls. For anyone familiar with the romance scam industry, that is enough to understand what is going on. For a naive applicant, it is not.

The legal side is highly ambiguous. A person who simply reads from a script may sincerely not realize they are taking part in fraud, especially if the employer carefully hides the operation's true goal. Even so, in a number of jurisdictions, participating in such schemes, even without intent, can be classed as complicity. Those who use someone else's face without that person's knowledge, or with formal consent but for illegal purposes, violate several areas of law at once: those covering biometric data, fraud and personal data protection.

The investigation shows plainly how AI tools are becoming infrastructure for criminal business. Romance scams with deepfakes are no longer isolated incidents but a systematized industry with recruiting, scripts, KPIs and division of roles. Victims lose money, unwitting "models" risk criminal prosecution, and platforms such as Telegram respond slowly to obvious violations. Until regulators work out a clear response and platforms introduce real controls, such schemes will only keep scaling.