Militant factions linked to Boko Haram in Nigeria’s northeastern region have reportedly begun integrating generative artificial intelligence tools into their operational strategies, according to a groundbreaking study published in 2026 by researchers at the Cambridge Programme on AI Science & Policy. The findings, based on 57 interviews conducted with 27 former members between 2025 and 2026, reveal a shift from traditional propaganda uses to more tactical applications.
Beyond propaganda: AI in combat planning and logistics
While AI-driven propaganda generation has long been associated with armed groups, the study highlights a broader range of applications. Former mid-level commanders and technical specialists within Boko Haram’s factions—ISWAP (Islamic State’s West Africa Province) and JAS (Jama’atu Ahlis Sunna Lidda’awati wal-Jihad)—describe how they now rely on publicly available AI platforms such as ChatGPT, Claude, Gemini, Grok, Meta AI, and DeepSeek to enhance their activities.
The integration of these tools reportedly began in 2023, shortly after their public release. Fighters have used AI to:
- Plan and refine attack strategies by analyzing past failures and testing hypothetical scenarios;
- Streamline logistics such as route planning, resource allocation, and unit coordination;
- Troubleshoot technical issues related to seized military equipment, weapon maintenance, or equipment malfunctions;
- Design improvised explosive devices and improve operational security measures.
When an AI system refused to respond to a query, fighters adapted by rephrasing prompts, altering contexts, or switching platforms—a workaround that allowed them to bypass restrictions more effectively.
Structured AI units: a Daesh-inspired model
The adoption of AI has taken a more institutionalized turn with the creation of specialized cells within Boko Haram factions. These units, composed of members skilled in engineering, armaments, explosives, and strategic planning, serve as on-the-ground technical support hubs.
Field operatives submit questions to these cells, which then query multiple AI systems before filtering and relaying the most actionable responses. Access to computers, accounts, and subscriptions is tightly controlled, restricted to trusted individuals to minimize risks of surveillance or data leaks.
This structured approach was reportedly encouraged by Daesh-affiliated instructors, who provided in-person training, remote assistance, and even equipment such as laptops, projectors, encryption software, VPNs, and paid AI platform access. Training covered AI fundamentals, prompt engineering, and techniques to circumvent platform restrictions. These skills were then disseminated across operational units.
Limitations and growing trust in AI
The study acknowledges methodological constraints: its conclusions stem largely from testimonies of former mid-ranking members, not current leadership, and lack independent verification of each reported use. Additionally, AI-generated advice may sometimes be inaccurate, incomplete, or impractical in real-world conditions.
Despite these caveats, the research underscores a critical shift: AI is no longer a theoretical risk but a tangible tool accessible even to resource-limited armed groups. While it does not replace human expertise, it accelerates knowledge acquisition and facilitates technical problem-solving across domains like translation, data analysis, maintenance, and planning.
For tech developers and security services, the challenge is twofold: detecting malicious use without stifling legitimate access to widely used platforms, and adapting surveillance mechanisms to counter organized, technologically adept adversaries rather than isolated users. The report calls for enhanced collaboration among developers, public authorities, security agencies, and researchers to strengthen safeguards against structured misuse.
As Boko Haram factions increasingly trust AI’s potential, they may invest further in subscriptions, computing hardware, and specialized training—raising a pressing question: to what extent can security frameworks evolve quickly enough to prevent AI from becoming a force multiplier in insurgent violence?