TFGBV affects millions of people, and the impact of TFGBV on women, girls, LGBTIQ+
persons, and other specifically exposed groups is profound and multifaceted. In a study from
2021, 60 percent of women in the Arab states reported experiencing some form of online
harassment.3 Globally, different studies indicate that between 16-60 percent of women are
affected.4 Looking specifically at women human rights defenders, activists, journalists and
media workers, the number is as high as 70 percent.5 Perpetrators are often associated with
anti-gender and anti-feminist individuals or groups, but can also be state agents or other users
of ICT.
This is why International Media Support (IMS), Jordan Open Source Association (JOSA),
KVINFO and Tadwein set out to create an AI-powered tool for gendered hate speech detection
in 2025, focusing on the Egyptian dialect of Arabic. The project has included collecting
experiences from a pilot project in Jordan, a robust annotation process, focus group
discussions with experts, training of trainers and many more elements, in order to produce an
AI-model with high precision of detection and intuitive useability. The model includes features
facilitating use for a range of tech savvy to non-tech interested activists, researchers,
journalists and organisations.