Members of Boko Haram and its affiliates are utilizing artificial intelligence chatbots from major technology companies to design more potent explosives, coordinate military operations, and resolve tactical challenges, according to new research from the University of Cambridge.
The study involved interviews with 27 former combatants from two factions: the Islamic State in West Africa Province (ISWAP) and Jama’at Ahl as-Sunnah lid-Da’wah wa’l-Jihad (JAS). Many respondents reported routinely consulting tools such as OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini, Elon Musk’s Grok, Meta AI, and China’s DeepSeek for technical and military guidance before, during, and after engagements.
Antonia Juelich, the Cambridge researcher behind the study, told Al Jazeera that approximately ten of the 27 interviewees discussed using AI in connection with explosives. This included efforts to enhance bomb power and recover explosive materials from unexploded military ordnance.
“There is nothing we haven’t received an answer on,” said one former ISWAP commander cited in the report. The research highlights a sophisticated, organized approach within the group. Operatives traveled to ISWAP-controlled areas in northeastern Nigeria to train senior members on AI usage. Fighters were gathered around laptops and projectors to learn how to query chatbots and circumvent safety safeguards.
“The white guys came and taught us,” one former commander recalled. “They assembled the top people in a room and used a projector to show how it works on a big screen.”
BOTH ISWAP and JAS reportedly established dedicated AI units staffed by specialists, including engineers and bomb-makers. One former JAS commander described a team of 23 operating from a solar-powered facility. Members analyzed information, answered queries, and trained senior leaders. “It is like a human robot. We used it a lot,” another former commander stated, adding that bomb-makers frequently kept laptops nearby to troubleshoot issues while assembling devices.
Juelich expressed concern that AI offers tailored, real-time problem-solving capabilities rather than generic instructions found in manuals. Fighters could specify available materials and receive customized advice for constructing weapons or handling unexploded munitions. In one incident, a disagreement over how to disarm a failed device led to an explosion that killed 40 people; AI is now consulted to mitigate such risks.
Beyond explosives, AI was employed across various combat stages. Combat footage was uploaded to ChatGPT for analysis and tactical instruction. When facing unfamiliar weapons, commanders directed fighters to “Just ask Grok.” AI also helped adapt commercial drones for bombing runs and guided motorcycle tactics to overcome government trenches, despite significant casualties during training.
Exclusive evidence reviewed by Al Jazeera further shows that supporters of the broader ISIL movement share guides on bypassing AI safety controls. Researchers from Tech Against Terrorism infiltrated a pro-ISIL technical forum where users discussed “jailbreaking” AI systems. One supporter likened the process to “removing the speed limiter of the car” to make it “more useful and faster than the manufacturer intended.”
Lucas Webber of Tech Against Terrorism described these discussions as merely “the tip of the iceberg.” The findings align with February warnings from United Nations officials to the Security Council regarding ISIL’s increased use of advanced AI. Researchers and tech industry leaders have cautioned that AI capabilities are outpacing safety safeguards.
Major AI companies assert they have measures in place to prevent misuse. OpenAI prohibits its systems from being used for terrorism or weapons development, while Anthropic, Meta, and xAI cite classifiers and other safeguards. DeepSeek and Google were also contacted for comment.
Anthropic CEO Dario Amodei has called for a slowdown in the development of advanced models to allow safety checks to catch up. In contrast, the US administration under President Donald Trump advocates for minimal regulation to foster innovation. Juelich noted that the Cambridge interviews focused on 2023 and 2024, raising concerns about the heightened risks posed by more powerful AI systems available in 2026.
“I worry about where this is heading,” Juelich said.
Leave a Reply