A new study from the University of Cambridge says fighters from Boko Haram factions have been using AI chatbots to help build bombs, plan attacks and solve problems on the battlefield.
Researchers spoke with 27 former members of two Boko Haram-linked groups. These are the ISIL affiliate ISWAP and the faction known as JAS.
Several former fighters said they routinely used tools from OpenAI, Anthropic, Google, Elon Musk's xAI, Meta and China's DeepSeek for technical and military advice.
How Fighters Say They Used the Technology
One former ISWAP commander said outside trainers came to teach senior members how to use the chatbots. They used laptops and a projector to show fighters how to ask questions and get around built-in safety rules.
The trainers reportedly provided laptops with privacy tools and set up paid accounts on multiple AI platforms.
According to the study, both factions created specialized AI units. One former JAS commander described a team of 23 people working from a solar-powered room, answering questions and training commanders.
Researcher Antonia Juelich, who authored the study, said roughly 10 of the 27 interviewees discussed using AI specifically for explosives. This included trying to make bombs stronger and pulling usable material from unexploded ordnance.
Juelich said the concern was not that this information exists elsewhere. It was that AI could respond directly to the exact problem a fighter had in front of them.
Fighters could describe the materials they had on hand and get tailored answers, then return with follow-up questions as problems came up.
Beyond Bomb-Making
The reported use of AI went further than explosives. One account described a fighter wearing a camera in combat, with footage sent back to a commander who analyzed it using ChatGPT and passed new instructions to the front line.
In another case, a fighter asked how to operate newly acquired firearms and was told to "just ask Grok."
Commanders also reportedly used AI to rethink troop numbers after losing 60 men in one operation involving around 200 fighters.
Some fighters said they consulted AI after failed attacks, describing their tactics and asking why an operation did not work.
Separately, researchers at Tech Against Terrorism found pro-ISIL supporters on an online forum discussing how to bypass AI safety controls. One user compared it to "removing the speed limiter" on a car.
Lucas Webber of Tech Against Terrorism said discussion of these bypass methods, often called jailbreaking, is increasing.
Major AI companies say they already have rules against this kind of use. OpenAI, Anthropic, Meta and xAI all told Al Jazeera they use safeguards meant to catch and block dangerous requests.
The interviews mostly cover AI use during 2023 and 2024. Juelich said she is concerned about what fighters could now do with more advanced AI systems available in 2026.
