The AI has already fallen into the wrong hands and they are using it to make bombs
Artificial intelligence has quickly become a go-to tool for everything from writing emails and summarizing meetings to helping students learn or engineers debug. But the same technology that saves people time can be used again, and a new report shows that terrorist organizations are finding ways to do just that.
According to a research paper shared with the New York Times ahead of its publication, researchers found evidence that Boko Haram members were using popular AI chatbots to support both day-to-day operations and combat-related tasks. Interviews with 27 former members conducted in Nigeria over the past two years suggest that tools such as ChatGPT, Gemini, Claude, Grok, Meta AI, and DeepSeek are used to gather technical information, solve weapons, and even help plan attacks.
This wasn’t just a few bad actors playing
What makes the findings particularly troubling is that this was not explained as the work of a minority of AI experimenters. The report says that group use of AI has become structured, with dedicated teams, internal training, and shared knowledge among members. The researchers also said some users were able to bypass built-in security protections designed to prevent the AI from responding to violence-related requests.
That’s not to say that today’s AI chatbots are giving dangerous instructions on demand. Most studies focus on operations until the end of 2024, and AI companies say that their new models include stronger protections. They also point out that many warnings fall into a gray area. For example, asking an AI how to fix a motorcycle or understand basic chemistry is not dangerous in itself, although the same knowledge can be misused.
The challenge now is to stay one step ahead
OpenAI told the New York Times that using ChatGPT to support terrorism or violence violates its policies and said it continues to improve its defenses against abuse. Meta similarly noted that the research involved mostly older versions of its AI models and said it has strengthened security measures since then. At the time, xAI and DeepSeek did not respond to requests for comment, according to the report.
Security experts also warn that AI is simply giving terrorist groups superhuman powers. Planning and executing an attack still depends on real-world implementation, funding, communication, and human interaction — things that a chatbot cannot replace. Instead, the researchers believe that AI is more likely to make less experienced members somewhat more powerful than to significantly change the threat landscape. Still, the report is another reminder that the AI security conversation isn’t just about cheating students on homework or people producing fake photos. As AI models become more powerful and accessible, companies will continue to face pressure to ensure that powerful tools remain useful to legitimate users while making it increasingly difficult for bad actors to exploit them.
