Strengthening Sudan’s information environment with AI

Sudan’s war is one of the world’s most dangerous conflicts, with millions of civilians affected by violence and displacement. It is also one of the riskiest operating environments for aid workers: more than 120 have been killed since fighting began

The information space is so polarised that any type of information – accurate or fabricated – can be weaponised and amplified by AI bots to incite violence. Humanitarian responders are targeted, communities are misled about availability of services, and aid delivery is obstructed when people need it most.

In several locations, narrative escalation has preceded or coincided with access constraints, reputational attacks, and protection incidents. Yet the humanitarian system has lacked a structured way to assess information-related risks, translate digital signals into operational guidance, or bridge international coordination with local Sudanese networks.

Our approach

Our pilot project responds to that need. Working with local Emergency Response Rooms and technology partner Valent, we’re using AI-tools to detect and analyse harmful information early, supporting Sudanese and international responders to counter dangerous narratives before they cause harm.

CDAC Network’s pilot project responds to this gap. Working with national NGOs, local Emergency Response Rooms, mutual aid groups and technology partner Valent, we’re building a model where technical detection, humanitarian interpretation and system response work together.

Valent detects coordinated inauthentic networks and narrative manipulation patterns. CDAC interprets these signals through a humanitarian lens, translating them into location-specific, protection-focused analysis. Together with Sudanese and international responders, we develop strategies to counter dangerous narratives before they cause harm.

This approach is distinctive because it:

  • Focuses on manipulation methodologies and network behaviour, not fact-checking individual content.

  • Translates technical outputs into operationally relevant, protection-sensitive guidance.

  • Remains politically neutral and grounded in humanitarian principles.

  • Embeds community actors throughout the analysis cycle.

Project focus

  • Early warning detection

    AI models detect inauthentic network behaviour online, analyse narrative amplification patterns and monitor their real-world impact – spotting threats humans might miss. 

  • Rapid response

    Responders in Sudan get up-to-date analysis to inform decision-making and develop counter-content strategies when harmful information emerges.

  • Community resilience

    Arabic-language resources help dispel myths, protect mutual aid group and Emergency Response Room volunteers, and reduce the panic and fear that harmful narratives aim to provoke.

  • Learning & adaptation

    Through CDAC’s Sudan Community of Practice, we create a space for local actors, UN agencies, NGOs and donors to share lessons, triangulate digital signals with ground realities and strengthen coordination mechanisms.

Co-designed with communities

Critically, these tools are co-designed with Emergency Response Room members and Sudanese NGOs to ensure they’re grounded in local expertise and actually work for the people who need them most.

By taking this ‘community-in-the-loop’ approach and embedding community actors throughout the analysis cycle, we strengthen the legitimacy and proportionality of outputs while avoiding a purely technical understanding of information risk.

Insights

  • Sudan’s information war: how weaponised online narratives shape the humanitarian crisis and response

  • Evolving dynamics of the Sudan conflict: implications for humanitarian action and civil society

  • Strengthening Sudan's humanitarian information environment with AI

  • Sudan: CDAC portal for responders