Boko Haram Uses ChatGPT, Claude and Gemini to Design Bombs and Plan Attacks

Boko Haram, the Nigerian terrorist organization responsible for years of violence in the Lake Chad region, has embedded commercial AI chatbots into its day-to-day combat operations far more deeply and quickly than previously believed. That is the finding of a new study published by the Cambridge Programme on AI Science and Policy, described on July 10, 2026 by the New York Times.

The study's author, Dr. Antonia Juelich, spent two years talking with former fighters in northeastern Nigeria. Her work overturns the earlier assumption that jihadist groups use artificial intelligence mainly for propaganda, and slowly at that. Accounts from former members point to the opposite: a systematic, organized rollout of AI for real combat tasks.

What the Chatbots Were Used For

According to the accounts gathered, fighters asked chatbots about designing improvised explosive devices, identifying captured military equipment, and troubleshooting weapons problems. The AI also advised on battlefield tactics and helped analyze attacks after they were carried out.

The most alarming element the study describes is the level of organization. Both main factions, classic Boko Haram and the Islamic State-linked Islamic State West Africa Province, created dedicated teams operating behind the front lines. Their job was to query AI systems, translate the answers into concrete guidance for commanders, and train further commanders in using the tools.

You type in a question, like how to build a bomb, and it tells you how. It's like a human robot. - former Boko Haram commander, quoted by researcher Antonia Juelich

Training From Foreign Fighters

The study found that knowledge of how to use AI did not develop independently within Boko Haram. Former ISWAP members described recurring training sessions led in person by foreign jihadist operatives between 2023 and 2025, during which selected commanders were taught to use chatbots for everyday combat and logistics tasks.

This shows that knowledge about circumventing AI safeguards is already being transferred between terrorist organizations on different continents, not just developed within a single group experimenting with a new tool.

Tech Companies Respond

OpenAI stated that the use of ChatGPT described in the study violates its usage policies and pointed to defense mechanisms it has since strengthened. Meta argues that the study covered older versions of its models and that safeguards have since been tightened. xAI and DeepSeek did not respond to requests for comment.

The case strikes at a sensitive point in the debate over AI safety. Companies have spent months touting progress in blocking requests for help building weapons or explosives, yet Juelich's study shows that an active terrorist organization still managed to embed these tools into its daily combat operations for several years.

Implications for Regulation

For European and Polish regulators, the case matters because the EU's AI Act introduces, starting in August, obligations for the most powerful models, including assessments of misuse risk. The Boko Haram case shows that even the strictest safety filters can fail against the determination of organized groups that share methods for bypassing safeguards.

The study does not state how many actual attacks were carried out with direct AI assistance, it focuses on documenting the scale and systematic nature of the tools' use rather than counting casualties. The author says further work is planned on how model providers could detect this kind of organized misuse before it becomes routine on the battlefield.

Sources: AI has already fallen into the wrong hands and they're using it to make bombs (digitaltrends.com), God has helped us, and so will AI: How the Terrorist Group Boko Haram Uses Frontier AI (askwhocastsai.substack.com)