A few years ago, Boko Haram fighters on motorcycles struck a Nigerian military outpost but were blocked by a surrounding ditch. The group paused, then consulted an AI system for solutions. A former commander described asking the tool how to clear such barriers after seeing similar scenes in films, providing details on their bikes and required distances. The AI supplied instructions for modifications. Mechanics adjusted the motorcycles for quicker speed and power. Riders built practice pits with hazards and rehearsed leaps, sometimes suffering injuries, until they could cross the gap and launch a successful raid. Research by Cambridge University scholar Antonia Juelich, based on interviews with former members, shows how generative AI now supports extremists in direct combat roles. Previously, groups like the Islamic State and al-Qaida mainly applied such tools for messaging, recruitment and basic security. Recent shifts include using chatbots to plan explosives, repair arms and develop attack strategies. Models have assisted across planning, execution and review phases. Safety features in these systems can be bypassed through repeated prompting. Juelich’s work and parallel studies note rising concerns over AI risks, including weapons development and terrorism support. Officials have compared advanced models to potent threats, though government focus remains on cyber vulnerabilities rather than militant applications. Experts observe that groups test multiple platforms to evade restrictions and continue adapting despite company updates to safeguards.
