Ledger comments on modified hardware wallets: what we know
Yesterday, we covered the reports of modified Ledger wallets. The company has now outlined the alleged attack scheme and announced new rules for device sales.
Published 10/17/2026Market Filter Editorial Team
AI Behind a Familiar Face: Fake Voices, Video Calls and Trust on Telegram
Open Measures examined discussions of voice cloning and face swapping in scam-linked Telegram channels. We explore the findings and the risk of trusting a channel author whose identity may be constructed with AI.
Photos, voice messages and videos can make us feel that we know the person on the other side of the screen. Participants in fraud schemes discussing artificial intelligence tools on Telegram are trying to exploit precisely that trust.
On October 16, 2026, Open Measures published research by Ryan Aliapoulios on channels associated with Yahoo Boys networks. The researchers found discussions of voice cloning, face swapping and fabricated images, alongside advertisements for paid training aimed at participants in such schemes.
What the researchers examined
Yahoo Boys is a term used for networks of West African cybercriminals known, among other activities, for romance scams and impersonation. In these schemes, perpetrators build relationships with their victims, then exploit that trust to obtain money or personal information.
Open Measures analysed posts published since the start of 2026 in a sample of Telegram channels and groups available in its datasets. The researchers searched for “yahoo” in channel names and manually checked the affiliation of certain channels.
The study highlighted four channels responsible for 6,629 posts matching the search query. This represents part of the activity observed, rather than a complete picture of these networks on Telegram.
How AI helps build a convincing identity
Posts in the sample discussed tools for several purposes:
- creating photos and building fictional profiles;
- generating and cloning voices;
- producing videos with swapped faces and synchronised lip movements;
- changing someone’s appearance during a video call;
- creating images of documents that could be used to simulate identity verification.
Open Measures observed increasingly sophisticated recommendations throughout 2026: from basic generators of short audio clips and videos to tools advertised as suitable for real-time conversations.
Services mentioned included DreamFace, ElevenLabs and Magicam.ai. The researchers describe how channel participants proposed using these products. This does not mean that the developers support such activities or that every advertised method actually works.
For a scammer, these tools may address a specific obstacle: a victim asks for a new photo, a voice message or a video call. Synthetic content offers a way to attempt to pass that check while maintaining a stolen or fictional identity.
Training in these schemes is also sold on Telegram
Some of the channels examined resembled service storefronts. Their administrators advertised paid lessons, “VIP classes” and AI tool packages, promising assistance with celebrity impersonation and other forms of deception.
The research therefore documents both discussions of technology and attempts to profit from teaching people to use it for fraud.
When trust develops in a channel’s author
At Market Filter, we suggest considering this risk beyond individual fake accounts. The following scenario is our editorial observation, rather than an established finding of the Open Measures study.
You may spend a long time following what looks like an ordinary personal channel, becoming familiar with its author’s manner, photos, voice and stories about their life. Gradually, familiarity with the content turns into trust in the person. Yet the identity behind that trust may have been created partly or entirely with AI.
There are several possible situations. A real author may use AI to help write posts, process their voice or create videos. A team may also run a channel built around a fictional character. A separate risk arises when a synthetic identity is used to impersonate a particular person or conceal deception.
Using AI does not, by itself, establish fraud. The relevant question is what readers are being asked to accept as real: assistance with content creation, an openly fictional character, or a supposedly real person with a personal history and experience.
A reader could therefore begin to trust an author, only to discover later that the familiar face, voice and biography were part of an artificially constructed identity.
The Open Measures study does not establish how often this happens in ordinary personal channels. However, the attempts it describes to fabricate signs of identity help explain the underlying risk: a recognisable persona and convincing communication do not establish who is behind them.
Our Methodology
Every channel in the catalogue is checked by the editors against one method: it is how we work out the rating, place a channel at a risk level and decide whether it belongs in Market Scam.