Terrorist Groups Leverage AI for Tactical Combat Advantages

A Boko Haram attack on a Nigerian military base using motorcycles was initially blocked by a defensive trench. After regrouping, the militants turned to artificial‑intelligence tools to learn how to overcome the obstacle.

They asked an AI chatbot for step‑by‑step instructions, feeding it details such as the type of motorcycle and the required jump distance. The model provided guidance that the extremists used to modify their bikes.

Mechanics upgraded the motorcycles for faster acceleration and higher speed. Riders dug pits filled with broken glass and fire, practiced jumps—some with fatal results—until they achieved sufficient lift to launch a successful assault.

Dr. Antonia Juelich’s research, shared with The New York Times ahead of publication, shows that generative AI is now assisting terrorist groups directly on the battlefield, extending beyond the propaganda uses that dominated earlier discussions.

Historically, groups like Islamic State and Al Qaeda relied on AI mainly for information‑operations such as propaganda, translation, recruitment and security tradecraft. Recent developments indicate a shift toward tactical, on‑the‑ground applications.

Built‑in safety limits on chatbots can be circumvented; researchers have found ways to coax models into providing restricted or harmful information through persistent prompting.

Dr. Juelich interviewed roughly 60 former Boko Haram members in Nigeria over the past year. The testimonies revealed that chatbots were used to design explosives, upgrade weapons and brainstorm attack methods.

Large‑language models, she writes, are now consulted at every stage of military activity—from mission planning through post‑operation analysis—marking a departure from earlier propaganda‑focused AI use.

The research arrives amid growing concern that advanced AI could pose threats comparable to “digital nuclear weapons,” with potential impacts on terrorism, biological weapons creation and other security risks.

The Trump administration has recently urged leading labs to allow government vetting of the most powerful platforms before public release, focusing mainly on cyber‑exploitation concerns rather than terrorist misuse.

“The terrorists are not waiting for us to make AI safe,” Dr. Juelich said, noting that the scope and nature of terrorist AI use have been “significantly underestimated.”

Daniel Byman of Georgetown University’s Center for Strategic and International Studies says terrorist groups are “mixing and matching” different AI systems to bypass technical guardrails. Boko Haram was platform‑agnostic, using ChatGPT, Claude, Gemini, Grok and DeepSeek interchangeably.

The described tactics mostly date to late 2024. AI companies claim newer models have stronger safety measures, and they note the dual‑use nature of many AI outputs—knowledge that can also serve legitimate purposes.

Some accounts were explicitly violent: a former Islamic State commander said an AI chatbot provided detailed bomb‑making instructions, calling it “like a human robot.”

Anthropic, Google, OpenAI and Meta responded to the findings, defending their products’ safety policies and asserting that harmful queries are generally refused. xAI and DeepSeek did not comment, and Pentagon officials declined to discuss AI‑enabled plots.

The nonprofit Future of Life Institute’s latest safety ratings show a decline across the industry, with xAI and DeepSeek receiving failing grades.

Other recent studies corroborate the Boko Haram findings. The CSIS report lists AI applications such as reconnaissance, translation, target research, IED design, itinerary planning, document drafting, coding, communications security and open‑source intelligence analysis.

Tech Against Terrorism’s testing program found that leading models fully refused malicious prompts only 57 % of the time. While explosives‑related queries were declined about 80 % of the time, improvised chemical‑weapon prompts were refused roughly one‑third of the time.

U.S. intelligence analysts also report that terrorists are beginning to use AI to 3‑D‑print weapon parts, drone components, repair parts and munitions fittings, according to a former senior official.

Experts caution that AI is unlikely to instantly transform terrorism. Terrorist organizations tend to adopt technology cautiously, selectively and pragmatically.

The collected testimonies highlight both eagerness and dedication. Defectors described organized training sessions—delivered via transnational jihadist networks—focused on leveraging generative AI, using VPNs and encryption, and evading safety restrictions.

These practices resemble corporate AI adoption, with teams dedicated to managing accounts, generating useful responses and navigating platform limitations, albeit for markedly different objectives.

Some analysts note that AI has so far played a larger role in lone‑actor, inspired attacks than in large‑scale operations orchestrated by established groups.

Aaron Zelin of the Washington Institute for Near East Policy cited recent research showing that suspected ISIS supporters in the U.S. and Europe asked ChatGPT for target information and attack methods, though no successful plots resulted.

Zelin highlighted the case of a 27‑year‑old Tunisian arrested in May for planning an attack on a museum or Jewish site in Paris with AI assistance.

The CSIS report also notes that AI can enhance terrorist financing by improving fraud and deception capabilities used to raise money for networks, member support, equipment and communications.

U.S. officials and researchers stress that operational constraints remain; AI is not expected to replace the trust, coordination, financing and real‑world experience that seasoned operatives rely on.

“The likely result is therefore not a dramatic increase in highly sophisticated attacks but rather a modest increase in the competence of lower‑level actors,” the study concluded.

Nevertheless, analysts warn of the technology’s reach. ISIS‑K, the group’s most virulent affiliate, has actively encouraged followers to use AI to evade detection by authorities.

Tricia Bacon of American University, a former State Department counterterrorism analyst, said, “AI has the potential—and in a few cases has demonstrated the ability—to accelerate the process of radicalization and mobilization to violence.”

Source link