Explained: Why AI's next major security challenge could be terrorism

For years, concerns over artificial intelligence focused on misinformation, cyberattacks and job disruption. Now, researchers and counterterrorism experts warn another risk is rapidly emerging: terrorist groups are increasingly using mainstream AI tools to support planning, logistics and battlefield operations, testing the limits of existing safety guardrails.

Artificial intelligence has transformed everything from workplaces to software development, but security experts are increasingly warning that its misuse by terrorist organisations could become one of the technology's most pressing challenges.

Fresh research by University of Cambridge researcher Antonia Juelich, alongside a new report from the Center for Strategic and International Studies (CSIS), suggests extremist groups are moving beyond using AI simply for propaganda or recruitment. Instead, they are increasingly turning to commercially available AI tools to assist with operational planning, tactical decision-making, reconnaissance and battlefield logistics.

More from Tech

The findings, based on interviews with former Boko Haram members and supported by the CSIS report and other counterterrorism studies, indicate that generative AI is becoming part of terrorist organisations' day-to-day operations. Researchers say the shift raises new questions about whether current AI safety measures are sufficient to prevent misuse as increasingly powerful models become widely accessible.

What has the new research found?

A forthcoming paper by Antonia Juelich, a terrorism and technology researcher at the University of Cambridge, is among the clearest indications yet that AI is becoming part of operational planning rather than simply a propaganda tool. Over the past year, Juelich interviewed nearly 60 former militants, including 27 former members of Boko Haram, to understand how the organisation had adopted AI technologies.

Her research found that large language models were being used before missions, during attacks and after operations to improve tactics, solve technical challenges and evaluate outcomes. Rather than being limited to recruitment or online messaging, AI had become embedded across multiple stages of military activity.

The findings are backed by separate reports from the Center for Strategic and International Studies (CSIS), Tech Against Terrorism and former US intelligence officials, all of which point to a broader trend of extremist groups experimenting with commercially available AI tools.

The CSIS report, authored by Daniel Byman and Seth G. Jones, reaches a similar conclusion but offers an important caveat. While AI is lowering technical barriers and making terrorist groups more efficient, the authors argue that extremist organisations remain "adopters rather than innovators." Rather than building cutting-edge AI systems themselves, they are expected to rely on commercial AI models, open-source tools and widely available software to improve existing operations.

How are terrorist groups using AI on the battlefield?

One of the most striking examples in Juelich's research centres on an attack against a Nigerian military base protected by a defensive trench. After an initial assault failed, Boko Haram fighters reportedly turned to an AI chatbot for help after recalling motorcycle jumps seen in films.

According to a former Boko Haram commander interviewed by Juelich, the militants entered details such as the type of motorcycle they used and the distance they needed to clear. The chatbot then suggested ways to improve the motorcycles' acceleration and speed.

Mechanics modified the bikes accordingly, while fighters repeatedly practised the jumps using makeshift obstacles. Some reportedly died during training before the group eventually succeeded in carrying out another assault.

Former militants also described using AI chatbots to help repair or upgrade weapons, design explosive devices, simplify technical instructions and generate ideas for attacks. Researchers say these examples show terrorists increasingly treating AI as a practical assistant capable of breaking down complex tasks into easy-to-follow guidance.

The CSIS report says AI can also assist with reconnaissance, translation, itinerary planning, coding, communications security, document drafting and analysing publicly available intelligence. Much of this information already exists online, but AI allows users to organise, personalise and retrieve it much faster, making complex tasks more accessible to less experienced operatives.

Which AI platforms are being used?

Unlike earlier assumptions that extremist groups relied on a single platform, Juelich found Boko Haram members used multiple AI systems interchangeably. These included OpenAI's ChatGPT, Anthropic's Claude, Google's Gemini, xAI's Grok and Chinese firm DeepSeek.

Daniel Byman, a terrorism expert at Georgetown University and co-author of the CSIS report, said terrorist organisations are increasingly "mixing and matching" different AI models to find responses that may bypass individual companies' safety restrictions.

The research also found that some militants attended dedicated AI training sessions delivered through transnational jihadist networks. According to interviewees, these sessions covered everything from creating AI accounts and writing effective prompts to using VPNs, encryption software and techniques for avoiding platform safeguards.

Are AI safety guardrails working?

Despite industry efforts, several recent studies suggest existing safeguards remain inconsistent.

The nonprofit Future of Life Institute recently concluded that safety commitments across much of the AI industry have weakened over the past year, with xAI and DeepSeek receiving failing grades in its assessment.

Meanwhile, Tech Against Terrorism, a UN-supported nonprofit, tested more than two dozen leading AI models using thousands of prompts based on real terrorism cases. It found that only 57 per cent of the prompts resulted in complete refusals. While requests involving explosives were rejected most of the time, prompts concerning improvised chemical weapons were blocked far less consistently.

The CSIS report also notes that while leading AI companies have introduced safeguards into their flagship models, thousands of open-source AI models remain freely available online. Many of these lack comparable guardrails, allowing extremists to experiment with AI systems that impose far fewer restrictions than commercial chatbots.

Former US intelligence officials have also warned that some extremist groups are beginning to use AI to support the design of drone components, repair parts and 3D-printed weapon accessories.

Why does this matter?

Counterterrorism experts caution against assuming AI will fundamentally transform terrorism overnight. According to the CSIS report, terrorist organisations have historically preferred reliable, low-cost methods such as firearms, improvised explosive devices (IEDs) and vehicle attacks over experimenting with risky new technologies.

The report argues that while AI lowers barriers by simplifying technical information and assisting with planning, it cannot replace the trust, organisational coordination, financing, operational security and real-world experience required to execute sophisticated attacks. Researchers therefore expect AI's biggest impact to be evolutionary rather than revolutionary.

Instead, AI is likely to make extremist organisations more efficient by improving propaganda, recruitment, disinformation, logistics and operational planning. It could also enable smaller groups and lone actors to perform tasks that previously required larger organisations or greater technical expertise.

At the same time, AI is expected to strengthen governments' counterterrorism capabilities. The CSIS report says intelligence agencies can use AI to analyse vast amounts of data, detect extremist content, identify suspicious financial transactions, track travel patterns, recognise behavioural anomalies and improve threat assessments far more quickly than human analysts alone.

Ultimately, researchers argue that AI is unlikely to produce a sudden surge in highly sophisticated terrorist attacks. Rather, its effects are expected to be gradual, steadily improving the capabilities of lower-level actors while simultaneously giving governments more powerful tools to detect and disrupt them.

That balancing act is becoming one of AI's biggest security challenges. As increasingly capable AI systems become widely available, governments, technology companies and researchers face the difficult task of preventing misuse without stifling innovation. The question is no longer whether extremist groups will use AI—they already are—but how effectively democracies can stay ahead of them while protecting civil liberties.