Researchers interviewed 27 former members of Boko Haram’s two factions, including the ISIL affiliate Islamic State West Africa Province (ISWAP) and Jama’at Ahl as-Sunnah lid-Da’wah wa’l-Jihad (JAS). Several interviewees said members had used AI systems including ChatGPT, Claude, Gemini, Grok, Meta AI and DeepSeek, Al Jazeera reported.
The reported uses extended beyond general information. Former fighters described using AI to address technical problems during operations, analyse battlefield situations and obtain advice related to weapons and explosives.
***
Antonia Juelich, the Cambridge study’s author, said that about 10 of the 27 people interviewed discussed AI use in connection with explosives. She said the most significant concern was the ability of AI systems to respond to specific circumstances rather than simply provide generic information already available online.
The research also found that ISWAP and JAS had established dedicated AI units involving members with technical and operational expertise. Former members said fighters were trained to use AI systems and, in some cases, were provided with laptops, accounts and paid subscriptions.
Al Jazeera also obtained evidence from a pro-ISIL online technical forum showing supporters discussing ways to bypass AI safety controls. Researchers from Tech Against Terrorism gained access to the forum and documented discussions about “jailbreaking” AI systems to obtain information that would normally be restricted.
Lucas Webber of Tech Against Terrorism said researchers were seeing increasing discussion about methods to circumvent AI safeguards.
The findings come amid broader international concerns about extremist groups exploiting increasingly capable AI systems. In February, United Nations officials warned the Security Council about the growing use of advanced AI by ISIL and its affiliates.
Major AI companies have policies prohibiting the use of their systems for terrorism, violence and weapons development. Companies including OpenAI, Anthropic, Meta and xAI have also introduced safeguards intended to identify and block dangerous requests.
The companies were contacted by Al Jazeera for comment on the findings.
Juelich said the Cambridge interviews largely concerned AI use during 2023 and 2024, making the rapid development of AI systems since then an additional concern.
She told Al Jazeera that researchers were concerned about what groups with prior experience using AI might be capable of doing with more advanced systems available in 2026.
The findings highlight an ongoing challenge for AI developers and regulators: preventing malicious use of increasingly capable systems while allowing legitimate users to access information needed for research, education and other purposes.
