Police patrol inside the Government Science school in Kankara, in northwestern Katsina state, Nigeria, after gunmen abducted students on 15 December 2020. Boko Haram claimed credit for the attack. (Photo by Kola Sulaimon, AFP, Getty)

Terrorist Groups Use AI as a Battle Consultant, Not Just a Propaganda Tool

By Claire Meyer

13 July 2026

Factions of Boko Haram are using leading artificial intelligence (AI) systems for every stage of military activity, from mission preparation to execution to post-mission review, according to new research from the University of Cambridge published on 10 July.

Terrorist use of AI is well documented, but primarily for the creation of propaganda and to help attract new recruits. The study by Antonia Juelich at the Cambridge Programme on AI Science and Policy (CASP), “God has helped us, and so will AI”: How the Terrorist Group Boko Haram Uses Frontier AI, found that Boko Haram members use AI as a problem-solver for bomb-building, weapons troubleshooting, tactical and strategic planning, operational security, and logistics.

In one case, operatives in the Boko Haram faction the Islamic State West Africa Province (ISWAP) used AI to work out how to jump defensive trenches with motorcycles to breach fortified bases. In another case, operatives used AI to learn how to maintain and use stolen specialized firearms.

Boko Haram is known to be one of the most adaptable violent non-state actors, with a high capacity for innovation, the report said. Since its emergence in the early 2000s, conflict involving the jihadist insurgency has killed approximately 43,000 people, displaced around 3.1 million, and left more than 4 million people severely food-insecure in the Lake Chad Basin region. The movement has close ties to the Islamic State (ISIS), and Boko Haram is a key node in ISIS’s global network.

The research found that ISIS operatives delivered in-person and online AI training for ISWAP across multiple locations starting in around 2023, and they supplied ISWAP with encrypted laptops, paid subscriptions, and advice on prompting techniques and bypassing platform restrictions over the course of several years.

“Because ISIS disseminates technical capabilities across its provinces and runs them like an integrated global network, similar training has likely reached other affiliates,” the report said.

Juelich interviewed former commanders in Boko Haram, who said that different factions have set up dedicated AI teams of between five and 20 people that query AI models to feed information into daily operations. They used the AI tools to coordinate better attacks using smaller units and to build bombs that reduce casualties in their own ranks. Some of the former Boko Haram members interviewed for the research said that they had used AI to craft better weaponized drones.

All of this indicates an uplift in threat actors’ capabilities—an improvement in capabilities or outcomes without requiring significantly increased resources. It could lower barriers to entry, increase the speed and frequency of attacks, increase the scale or sophistication of attacks, improve precision, and improve operational security.

In addition, the benefits play off each other. AI use to improve recruitment can help with targeting operatives with aviation knowledge, which can then accelerate intelligence gathering on airport security protocols, enabling streamlined procurement of drone components and boosting secure communication during operations planning and enactment. These parallel improvements “enabled complex operations, such as coordinated multi-site drone attacks, that would be impossible with any single enhancement alone,” the report said.

Those AI teams are also adept at skirting around the platforms’ safeguards. While advice on vehicle repair or logistics wouldn’t clash with AI guardrails, requesting guidance on explosives design does. But one commander told researchers that the restrictions were manageable, not prohibitive, and that users could bypass restrictions by explaining they needed the details for a movie or something similar. Operatives will also mix and match AI tools to get the answers they seek, so even if one model refuses to answer, another can likely be manipulated into providing actionable information.

The AI use cited in the CASP study is primarily from 2024, and AI companies claimed in a New York Times article published 10 July 2026 that their models’ safety measures have improved significantly since then. But that’s debatable. A Future of Life Institute study graded major AI firms on safety commitments last week and found they had mostly eroded since 2025. In March 2026, the Center for Countering Digital Hate released research showing how AI chatbots could be manipulated into providing information to enable mass shootings and other violent attacks. AI tests by Tech Against Terrorism found that AI models refused to answer prompts drawn from real-world terrorism cases just 57 percent of the time.

“Artificial intelligence is unlikely to transform terrorism overnight, analysts and U.S. officials say,” the Times reports. “Terrorist organizations typically adopt technology cautiously, selectively, and pragmatically. But the testimonials that Dr. Juelich collected depict both eagerness and dedication among Boko Haram cells.”