AI is now helping terrorists do Evel Knievel-style motorcycle jumps over defensive trenches… by practicing in homemade pits filled with broken glass and fire. Yes, really.

In this episode of Based Camp, Simone and Malcolm Collins break down the bombshell research from Cambridge University’s Dr. Antonia Juelich on how Boko Haram (and other groups) are actively using frontier AI models — including ChatGPT, Claude, Grok, Gemini, Meta AI, and DeepSeek — for real battlefield operations.

Former members describe getting step-by-step AI guidance on attack planning, crossing obstacles, building and improving bombs/explosives, weaponizing drones (including grenade carriers), troubleshooting weapons, logistics, vehicle repair, operational security, VPNs/encryption, and even jailbreaking safety filters by claiming the requests are “for a movie.”

We cover how AI has moved from propaganda and recruitment into day-to-day combat support, why low-skill fighters are getting the biggest capability jump, the failure of current AI “safety” theater, and what this actually means for national security.

Plus the usual unfiltered Based Camp takes: high-trust vs low-trust societies, why serious monitoring (outside the home) is non-negotiable, the case for ruthlessly eliminating existential threats instead of “managing” them, techno-Puritan futures, tech-lord breakaway states, the Israel alliance, and why the West keeps failing at this.

Show Notes

Thanks to Bruno and @NotAldousHuxley for this one!

Recently Antonia Juelich took to X to promote her research on this topic, which was covered by the NY Times: “In a hotel room in northeast Nigeria, I opened a leading AI chatbot, turned my laptop toward a former Boko Haram commander, and asked if he’d used it. He nodded.

“You type in the question… like ‘How can I build a bomb?’, and then it tells you how. It is like a human robot. We used it a lot.” she wrote.

She added: “Members consistently reported benefiting from using AI.

“Trial-and-error can kill you. AI gives you accuracy.”

“Anytime they didn’t understand something, they would ask the AI.””

The NY Times Article

NY Times: How Terrorist Groups Are Using A.I. to Gain an Edge in Battle

(Archive Link)

“When a gang of motorcycle-riding members of Boko Haram attacked a military base in eastern Nigeria a couple of years ago, they were stymied by a defensive trench surrounding the complex.

The extremists regrouped. Before launching another assault, they asked A.I. for help.

“We saw in a movie how motorcycles can jump over bridges,” a former Boko Haram commander told Antonia Juelich, a terrorism and technology researcher at Cambridge University. “We used A.I. to learn how to do this. We gave it information, like what motorcycles we use and the distance we need to jump and so on, and it gave us steps on what we have to do.”

Using tips from chatbots, mechanics modified the motorcycles to allow for faster acceleration and top speed. The riders dug their own holes, filled them with broken glass and fire, and practiced jumps — sometimes with fatal outcomes — until they achieved enough aerial liftoff to mount a successful attack, defectors said.

The episode, recounted in a research paper by Dr. Juelich shared with The New York Times ahead of its publication on Friday, highlights how generative artificial intelligence tools are increasingly aiding terrorist groups directly on the battlefield, experts say, despite efforts by their makers to safeguard them from misuse.”

The article covers a research paper recently published by Dr. Antonia Juelich of the University of Cambridge, who interviewed 27 Nigeria-based former Boko Haram members in 2025 and 2026 and found:

* Terrorism use of AI has expanded from just intelligence, propaganda, and recruitment to combat and day-to-day operations, including

* Weapons troubleshooting

* Designing explosive devices

* They’re circumventing the AI safety protocols

* They use ChatGPT, Claude, Gemini, Grok, Meta AI, and DeepSeek

* Islamic State operatives are delivering in-person presentations, like with projectors and laptops

The report: ““God has helped us, and so will AI”: How the Terrorist Group Boko Haram Uses Frontier AI”

AI Use Cases

Core combat and attack use cases

* Planning and optimizing attacks, including tactics for assaults and raids at different stages from preparation to execution to post-mission review.

* Using AI to work out how to cross defensive obstacles, such as calculating how to jump military trenches by motorcycle to breach fortified bases.

* Generating tactical guidance for “very tough events,” such as how to respond when military forces attack their positions.

* Receiving continuous AI-derived strategies from specialized units, which are then implemented by field commanders.

* Providing real-time or near real-time advice to commanders in the field via remote communication with AI units based in camps.

Weapons, explosives, and drones

* Asking step-by-step guidance on how to build bombs and other explosive devices.

* Designing more powerful explosive devices, including advice on explosive composition and payload parameters, to increase lethality while reducing their own casualties.

* Troubleshooting malfunctioning weapons, including servicing and repair guidance for firearms and other systems.

* Using AI to assist with the weaponization of drones, including advice on payload weight and release-mechanism design for grenade‑carrying drones.

Logistics, maintenance, and “dual‑use” support

* Obtaining vehicle repair guidance for cars, trucks, and motorcycles used in operations.

* Getting logistics advice on how to move supplies, plan routes, and manage basic procurement tasks.

* Drawing on general technical knowledge (engineering, electronics, etc.) to solve equipment problems that would otherwise require trial‑and‑error experimentation.

Operational security and communications

* Using AI to improve operational security practices, including advice on evading surveillance and countermeasures.

* Receiving guidance on secure communication, including the use of VPNs, encryption tools, and account compartmentalization.

* Getting help to phrase or structure content (e.g., pretexts like “this is for a movie”) to bypass model safeguards when asking for disallowed instructions.

Organizational management and internal training

* Using AI as a general “problem‑solver” and advisory tool for military and strategic decision-making by senior commanders.

* Relying on AI units to research solutions to specific technical or tactical problems submitted by field units, then turning model outputs into written or oral briefings.

* Employing AI in internal training workflows, with specialized units using AI themselves and then teaching mid‑ and upper‑level commanders how to prompt and apply outputs for their own units.

* Managing multiple premium AI accounts across different providers, including account setup, subscription management, and tool selection as an operational function of the AI units.

Propaganda, recruitment, and broader ecosystem (outside Boko Haram core but in the report)

* Jihadist supporters (as distinct from formal organizations) using AI to generate, translate, and adapt propaganda content, including AI‑generated news bulletins and images.

* Pro‑ISIS media and tech‑support outlets producing AI‑assisted guides on how to safely use AI tools for propaganda and information operations.

* Early experimentation by supporters with using AI as a digital advisor for movement building, including detailed planning steps for supporting or establishing a “caliphate.”

How are they circumventing safeguards?

* VPNs

* Structured jailbreak training

* Deceptive prompt framing

* Redundancy across accounts and models

* Former members report that “boys that have received extensive training … bypass the restrictions,” indicating a small cadre of more technically capable prompt engineers inside the group.

* One explicit ploy mentioned is telling the model that dangerous instructions are needed “for a movie or something like that,” i.e., wrapping malicious requests in a fictional or benign cover story.

What can be done?

* The Trump administration has tried to get leading labs to let them vet new models before they’re publicly released (The Trump administration forced the AI lab Anthropic to disable and restrict foreign access to its advanced Claude models, including Fable 5 and Mythos. The administration also asked OpenAI to delay the public rollout of its GPT-5.6 series amid similar national security and cybersecurity concerns)

* Per the NY Times:

* The major AI companies tried to defend themselves:

* “Asked about the Boko Haram study, Michael Aciman, an Anthropic spokesman, said the company’s products were “built to refuse dangerous requests, including those tied to violence, attack planning and building explosives.” He added that Anthropic worked with outside experts, researchers and industry partners because “no single company can counter these threats alone.””

...