Cambridge Study Says Terror Groups Are Misusing Major AI Chatbots

A study by researcher Antonia Jülich of the Cambridge Programme on AI Science & Policy says Boko Haram uses popular AI chatbots including ChatGPT, Claude, Gemini, Grok, Meta AI, and DeepSeek. The article reports that Jülich conducted 57 interviews with 27 former members of the group. According to the study, both Boko Haram factions have established dedicated AI units.

The Decoder reports that the group uses AI for attack planning, building more powerful explosive devices, weapons maintenance, and operational security. ISIS liaisons have reportedly trained commanders on bypassing safety filters, with prompt engineering and jailbreak training offered since 2023. The article also describes an ISWAP case involving AI-assisted attempts to copy motorcycle jumping techniques from a movie, in which 18 fighters died during training and eight made the jump.

The central policy concern is not just that chatbots can answer risky questions, but that safety filters reportedly failed to reliably prevent misuse. The article argues that voluntary self-regulation by AI providers is not enough if motivated users can repeatedly work around safeguards. It also notes that Anthropic has said jailbreaks will likely never be fully eliminated.

The article distinguishes general-purpose chatbots from more specialized AI systems. Researchers cited by The Decoder say chatbots mostly make existing knowledge easier to find, while specialized life-science AI systems could present a larger misuse risk. The key takeaway for readers: AI safety debates are moving from abstract speculation to documented misuse cases and governance questions.