HerWILL has successfully concluded the “AI for Digital Safety Datathon 2026”, a global initiative that brought together participants from Egypt, Bangladesh, Syria, Senegal and other countries to address the growing challenge of harmful online content.

The competition recognised three winners from a competitive pool: Azra Tuni from Bangladesh as champion, Nour Mohammad from Syria as first runner-up and Syed Zayan Anwar from Bangladesh as second runner-up.

They were selected from the top seven finalists representing Bangladesh, Syria, Egypt and Cameroon, demonstrating strong technical skills alongside a clear understanding of real-world digital safety challenges.

As digital spaces continue to expand, organisers said the need for responsible and context-aware AI systems has become increasingly urgent. The datathon focused on developing solutions to detect harmful online content, including explicit, subtle and neutral forms, particularly in multilingual and low-resource contexts.

The programme followed a structured pipeline, beginning with a seven-day workshop series covering machine learning, natural language processing and applied deep learning. Participants then applied their learning in a competitive environment using a real-world, human-annotated dataset.

Designed and executed by STEM Lead Saiyara Mahmud and Co-Lead Tanmoy Mazumder under the supervision of HerWILL Founder and CEO Farhana Hasan, the initiative engaged more than 2,167 participants across over 45 countries, 60 institutions and multiple cities worldwide.

The programme placed emphasis on detecting subtle toxicity beyond binary classification, addressing code-mixed language challenges across English, Bangla and Banglish, and evaluating models through a balanced approach using the F1NOP metric.

It also contributed to the development of a large-scale multilingual dataset of nearly 60,000 annotated samples, supporting future research in AI safety.

The initiative was led by HerWILL in collaboration with a global network of partners. Strategic partners included the National Science Foundation (NSF) and the AI Institute for Advances in Optimisation (AI4OPT). Premium partners included SmartBizPay, Build Bangladesh, Galaxy for Development and AMLD.

The programme was further supported by UIU Data Science Club, IUT Career and Business Society and Government Titumir College Data Science Club, with professional development support from BYLC. The Business Standard served as the print and media partner.

Organisers said the initiative demonstrates how inclusive and globally connected learning models can translate into meaningful AI solutions, helping build a generation of practitioners equipped to design safer digital systems.