Bellingcat Releases Machine-Learning Classifier to Surface Civilian-Harm Posts on Telegram

Investigative outlet Bellingcat said on June 25, 2026 that it has built a machine-learning model to detect and rank civilian-harm cases from Telegram posts, and published the training code notebook for it. The effort aims to ease the burden of manually screening large volumes of social-media posts and shift researchers' time from searching to verification, with details laid out in the official article.