Boko Haram exploited US and Chinese AI chatbots for attacks: Cambridge study

South China Morning Post

Boko Haram militants in Nigeria actively exploited American and Chinese artificial intelligence chatbots, including OpenAI's ChatGPT and DeepSeek, to assist in bomb construction and tactical attack planning throughout 2024 and mid-2025. This operational integration represents the first documented on-the-ground evidence of an active terrorist organisation utilizing frontier generative AI models for direct military preparations.

According to a Cambridge University study by researcher Antonia Juelich, specialized trainers from the Islamic State network instructed the group on bypassing safety protocols using prompt engineering and virtual private networks. The militants established dedicated internal AI units staffed with technical personnel to systematically jailbreak both open-source and proprietary platforms, rendering standard corporate safeguards ineffective against sophisticated exploitation. These revelations have intensified pressure on Washington and Beijing to address non-state AI misuse during their upcoming safety and security talks, particularly concerning the proliferation of easily modified open-source models amidst fierce US-China tech competition globally.

Comment

Transnational terrorist networks are increasingly acting as force multipliers by distributing technical exploitation methodologies across geographically dispersed affiliates. This democratisation of jailbreaking techniques bypasses traditional state-centric export controls and corporate content filtering. For Indian security agencies monitoring theatres like Jammu and Kashmir, this trend necessitates a shift from tracking physical supply chains to intercepting digital instruction sets. Counter-terrorism doctrine must therefore evolve to treat generative model vulnerabilities as active cyber-physical threat vectors.

To read in details click here