Eva Yayi Mawa, Co-Founder and Executive Director of Go Girl ICT Initiative - Courtesy

JUBA, South Sudan (Eye Radio) – In an increasingly digital world, artificial intelligence (AI) systems are rapidly becoming the gatekeepers of online safety. Yet, for nations like South Sudan, these high-tech guardians may be blind to the dangers lurking in local languages.

Eva Yayi Mawa, Co-Founder and Executive Director of Go Girl ICT Initiative, sounded the alarm this week during the discussion in Juba from June 18–20, 2025. She highlighted the urgent need to make AI systems inclusive of South Sudan’s vast linguistic diversity.

The workshop was part of a broader stakeholder consultation on “Enhancing Legal and Non-Legal Measures Against Hate Speech,” organized by the Union of Journalists in South Sudan in recognition of the International Day for Countering Hate Speech.

“You can write hate speech in your local language and the system won’t understand it or block it,” Eva told participants, who included young people, content creators, media professionals, ICT experts, CSOs, and government representatives.

South Sudan, with over 64 local languages (some estimates say over 70), faces a unique challenge.

Global AI tools used by platforms like Meta (Facebook) are primarily designed to detect offensive content in dominant languages such as English, French, Spanish, and a few African languages like Yoruba or Kiswahili.

However, South Sudanese languages remain largely invisible to these sophisticated systems.

Eva warned that this linguistic exclusion creates a dangerous loophole, allowing harmful content to thrive, particularly in conflict-prone regions where hate speech can easily inflame ethnic tensions.

At the same time, she emphasized that AI is a double-edged sword: while it accelerates the spread of online hate, it also holds immense potential for detecting and countering such content, if properly trained.

“AI systems are only as good as the data we feed them,” she explained. “We need ethical developers from South Sudan involved in training AI using our local dialects.”

Her organization, Go Girl ICT Initiative, is leading research under a project called YoGage AI, which specifically focuses on Juba Arabic, the country’s widely spoken lingua franca.

However, they have encountered significant challenges, notably dialectal variation and inconsistent pronunciation.

Eva noted, “Even within Juba Arabic, different communities pronounce the same words differently. This makes it hard to standardize a dataset for training AI.”

While some progress has been made in digitizing Nuer and Dinka language models, Eva underscored that a concerted national effort is needed to develop a comprehensive language library for natural language processing (NLP).

She also stressed the critical need for digital literacy, urging South Sudanese social media users to educate themselves on how these AI systems work and how to detect patterns of hate speech.

The workshop concluded with a strong call to action: build linguistically inclusive AI systems, invest in local tech talent, and ensure that no language is left behind in the fight against online hate.

This growing conversation highlights a critical digital divide where technological progress currently fails to meet linguistic realities, leaving communities at risk of being silenced or harmed by the very systems meant to protect them.