Violent extremist groups in West Africa have moved beyond using artificial intelligence for propaganda and recruitment, instead utilizing AI chatbots for direct operational support, including the design of explosives and the planning of military assaults.
The findings were detailed in a July 10, 2026, report by The New York Times, based on a year of field research by Antonia Juelich, a terrorism and technology researcher at the University of Cambridge. Juelich conducted nearly 60 interviews with 27 former members of Boko Haram and its splinter faction, the Islamic State West Africa Province (ISWAP).
Operational Use and Battlefield Application
Research indicates that the use of generative AI by these groups has shifted from creating slicker videos and cloned voices to reconnaissance, coding, weapons troubleshooting, and battlefield planning. Former commanders described using chatbots as “knowledgeable assistants” available at any time to provide detailed, step-by-step answers via text or voice input.
One documented case involved an assault on a Nigerian military base where fighters were unable to cross a defensive trench. After seeing motorcycles jump obstacles in a film, the group provided AI chatbots with the distance they needed to cover and details of their available motorcycles. The AI generated a set of steps for modification. Mechanics adjusted the bikes for increased speed and acceleration, and fighters rehearsed the jump—sometimes fatally—before successfully using the maneuver in a subsequent attack.
Beyond mobility, the groups have used AI to refine explosive devices. One insurgent noted that AI systems identified chemical combinations that created more powerful blasts than previous trial-and-error methods. Juelich’s research further states that AI has assisted in attack planning and the design of explosive devices.
Institutionalization and Tool Integration
The investigation revealed that these groups are not loyal to a single provider, treating various AI platforms as interchangeable tools. Former members cited the use of:
- ChatGPT
- Claude
- Gemini
- Grok
- DeepSeek
- Meta AI
Evidence suggests this is not isolated experimentation. According to Juelich, both Boko Haram factions have established dedicated AI units. Furthermore, Islamic State operatives have reportedly provided remote assistance and in-person training on how to use these tools.
Circumventing AI Safety Safeguards
While major developers like OpenAI, Google, and Anthropic implement safeguards to reject requests for dangerous content, former fighters claimed they routinely bypassed these restrictions. One method involves disguising prompts as legitimate hobbyist, engineering, or academic projects.
A study by the UN-backed non-profit Tech Against Terrorism analyzed more than 20 AI models and found that only 57% of thousands of terrorism-related questions were completely rejected. While approximately 80% of questions regarding explosives were rejected, only about one-third of questions concerning improvised chemical weapons were blocked.
In response to these findings, an OpenAI spokesperson stated the company will continue strengthening defenses against bad actors. Anthropic stated its products are designed to reject dangerous demands and that it works with external experts and partners. Google stated its policies prohibit using AI to cause real-world harm and claimed that, after reviewing content, the answers provided were not specific enough to lead to misuse.
Global Security Implications
The shift toward operational AI use has raised alarms among security officials. CIA Director John Ratcliffe has described advanced AI models in terms evoking weapons of mass destruction. AI safety researchers have argued that terrorism assistance and biological weapons design are underacknowledged risks compared to concerns over job displacement or disinformation.
While Daniel Byman, a terrorism expert at Georgetown University, cautioned against overstating the immediate transformation, he and other researchers flag the trajectory as requiring urgent attention. Juelich stated that the scale and nature of AI use in terrorism is “severely underestimated” and urged policymakers and developers to remain vigilant.
Find more reporting in our Tech section.
Worth a look