Cambridge Study: Boko Haram Systematically Uses ChatGPT, Claude, Grok for Attacks

A report from the Cambridge Programme on AI Science & Policy, authored by Antonia Juelich, argues that terrorist adoption of frontier AI is further along and more organized than earlier analysis assumed. The findings draw on semi-structured interviews with 27 former Boko Haram members conducted in northeast Nigeria across 2025-2026, covering AI-assisted activity that ran primarily through 2024. Rather than casual experimentation, the study describes formalized adoption inside the group — specialized units and internal training programs built around commercial chatbots.

The militants reportedly drew on a broad slate of mainstream models, including ChatGPT, Claude, Gemini, Grok, Meta AI, and DeepSeek, and applied them to combat operations, routine organizational tasks, attack planning, weapons troubleshooting, and explosive device design. The report also states that operatives managed to circumvent some vendor safeguards, and that Islamic State members passed AI know-how to Boko Haram in person through transnational jihadist networks — a knowledge-transfer channel that spreads capability faster than any single group could develop it alone.

The significance for the security and AI communities is the claim that misuse has moved from theoretical concern to observed practice at an operational scale. The authors frame it as a present and growing reality that demands attention from policymakers, security professionals, and the model developers themselves, whose guardrails were evidently bypassed. Notably, the evidence rests entirely on former-member interviews rather than technical forensics, so readers should weigh the recall-based methodology when assessing specific capability claims.