Boko Haram and ISIS Exploit ChatGPT and Gemini for Attacks

Boko Haram and ISIS Exploit ChatGPT and Gemini for Attacks

Terrorist groups are using every major AI chatbot for attack planning and weapon development.

ISIS is reported to have started offering training in prompt engineering and security bypassing since 2023, training Boko Haram commanders in Nigeria on how to circumvent AI security filters.

Boko Haram is now utilizing popular AI chatbots such as ChatGPT, Claude, Gemini, Grok, Meta AI, and DeepSeek. According to a new study by researcher Antonia Jülich from the Cambridge Programme on AI Science & Policy (CASP), both factions of the group have established dedicated AI units. Jülich conducted 57 interviews with 27 former members of the group.

The study revealed that the group uses AI for:

• Attack planning
• Manufacturing more powerful explosive devices
• Maintaining weapons
• Operational security

ISIS affiliates have trained commanders on how to bypass security filters. Both factions of Boko Haram have created their own dedicated AI units. According to Jülich, they "gathered the best elements in a room and used a projector to show how it works on a large screen."

In a particularly striking case, the ISWAP faction used AI to replicate motorcycle jump techniques from a movie so that fighters could cross trenches. Eighteen fighters died during the training, while eight successfully made the jump.

A former ISWAP commander (Munzir) tells researcher Antonia Jülich how the group used AI to learn how to jump over trenches on a motorcycle.

But of course, it doesn't stop at motorcycle stunts. "Former members described a strong enthusiasm for AI, and some stated that the group had previously considered weapons of mass destruction," writes Jülich. "Although Boko Haram's use of AI remains conventional, this should serve as a warning to take seriously the risk that terrorists seek AI assistance for chemical and biological weapons."

Researchers, as well as AI labs like OpenAI and Anthropic, have long warned that AI models could make dangerous knowledge more accessible. However, voluntary self-regulation does not seem to be sufficient, as the study revealed that security filters fail to reliably prevent abuse. This may simply be the nature of large language models: Anthropic recently admitted that security bypasses are unlikely to ever be completely eliminated.

Researchers also emphasize that chatbots like ChatGPT and Claude, while they have attracted public attention, are not the real concern, as they primarily facilitate the search for existing knowledge rather than generating something new. The greatest worry lies in the potential for abuse of more specialized AI systems in the life sciences.