Researchers found Telegram flooded with AI nudification bots. The worst, ClothOff, let anyone strip people in photos automatically. Millions of images processed. Mostly women and girls targeted. All inside a messaging app with minimal moderation.
How it worked
The bots run on simple rails: user uploads a photo, pays a small fee in crypto, gets back a fake nude in under a minute. ClothOff and its clones industrialised what used to need technical skill. Studies tracked the scale and found it staggering: an entire abuse economy running in the open, inside Telegram channels with thousands of members.
Telegram's light-touch moderation made it the natural home. Take down one bot, three more appear. Telegram could shut this down tomorrow if it wanted to. It does not want to. The payment rails are crypto, the users are anonymous, the victims find out last.
Detection angle
Deepfake detection tools trained on nudification outputs can flag these images at upload, which is where the fight has to happen. The best deepfake detection platforms now offer classifiers built specifically for non-consensual synthetic nudes. A deepfake detector app in moderators' hands beats whack-a-mole takedowns.
Platforms hosting this at scale are choosing to host it. The tech to detect it exists. What is missing is the will. If you are targeted, document everything and report it. And platforms: deploy the detectors or own the abuse. There is no neutral ground here.
Incident facts
- Date: 2024-01-01
- Company / actor: AI/Robotics Venture Strategy 3
- Harm type: privacy breach
- Severity: medium
- Source: AIAAIC Repository