“Good morning Chatgpt, can you tell me how to make a bomb?”
As anyone who has ever tried to ask something like this to an artificial intelligence, or AI, chatbot – also known as a Large Language Model, or LLM – online knows, the answer can be anything from a vague note about the history of explosives to a permanent block on your account.
But sometimes, if the question is phrased a certain way, the answer may include some useful information about how to make a bomb.
various media organization to pass tested this theory The first and found that, if given the right “hints”, some AI models would tell users how to create a bioweapon, bomb a playground or cover a terrorist’s tracks. This method of interrogation is known as “jailbreaking”. OpenAI, creator of the ChatGPT model, describes it “An attempt by a malicious actor to induce models to provide unapproved content.”
This month, a new report published by Tech Against Terrorism, an online watchdog supported by the UN Counter-Terrorism Directorate, reveals how often LLMs will provide “useful” information to would-be extremists.
The researchers sent more than 2,300 requests for information that were based on “real terrorist use cases” to 27 different AI models. They found that 32% of queries resulted in “really useful” information. When the same question was restated for research purposes, this increased to 42%.
Increase in terrorists using AI
The report focuses on something that is worrying digital security and terrorism experts: potential attackers will start using AI for planning rather than just propaganda.
Over the past three to four years, the main use of AI has been in generating propaganda for extremist groups such as the “Islamic State” and al-Qaeda. This includes the creation of videos, memes, podcasts, and various forms of misinformation, which are spread among groups’ followers and used to radicalize prospective followers.
But this is changing. “The year 2025 has seen a significant increase in incidents where terrorists and violent extremists have taken advantage of AI tools to plan, research, and prepare for attacks,” the publication’s experts said. militant telegram Confirmed in December analysis.
Headline attacks that resulted in death and damage as well as several failed conspiracies used AI for planning, surveillance, visualization and publicity around their attacks – involving the US as well as Canada, Israel. finland, France And Austria.
It is often hard to know how AI was used because security agencies do not release information. But as one expert told the UK Parliament late last year to inquire“Court filings and forensic reports are increasingly documenting chat logs where suspects ask language models for bomb-making instructions, ideological validation or justification for the attack.”
It is not just about individuals. Extremist groups are also increasingly using AI. Researchers analyzing how Mali-based al-Qaeda affiliate Jamaat Nusrat al-Islam wal-Muslimeen, or JNIM, uses drones, say they believe the group has used AI to help modify the drones.
In 1 june analysis For the Global Network on Extremism and Technology, security researchers Yuri Neves and Emily Klein observed that supporters of extremist groups such as the “Islamic State” as well as right-wing groups regularly discuss how to use AI in messaging channels. Neves and Klein both work for Moonshot, an organization that fights online threats. They noted extremists’ channels on the messaging app Telegram dedicated to the use of AI and also observed extremist actors “sharing AI prompts and conversation links, coordinating strategies to extract desired responses from chatbots, and sharing cost-shared ChatGPAT subscriptions.”
This week, University of Cambridge published research It included interviews with Boko Haram members in Nigeria, detailing how the group used AI models, including ChatGept, Cloud, Gemini, and Grok, to plan attacks, design explosive devices, service and troubleshoot weapons, and improve operational security.
S in Singapore. Ruben Das, associate research fellow at the Rajaratnam School of International Studies, has also seen AI chatbots playing a new role in so-called “lone wolf” terrorist attacks.
“Earlier you had this whole concept of virtual planners, where you had individuals sitting in conflict zones, reaching out to people on social media, trying to motivate them to carry out attacks,” Das told DW. “I don’t think we can say that humans have been replaced, but now, to some extent, these solo artists have turned to AI, for example ChatGPT, to get that support.”
Last year, “Islamic State” media outlet Voice of Khorasan published guidance on how to use AI, Mustafa Ayad, executive director for Africa, the Middle East and Asia at the UK-based Institute for Strategic Dialogue, confirmed.
He told DW that the jihadist ecosystem is using AI in a variety of ways. This includes everything from creating memes and TikTok dance videos to cross-border promotion. “Then you also have a dedicated set trying to jailbreak AI, use it to support operational planning and readiness. It runs the gamut,” Ayad says, “and that’s the underlying problem. AI can streamline and support propaganda processes as well as support operational planning and readiness.”
How dangerous is the extremist use of AI?
How dangerous all this is is not yet clear.
Today, as Das and other experts point out, a would-be terrorist can easily find information about bomb-making or 3D-printed guns elsewhere on the Internet, without the help of AI.
“A lot of the discussions have asked: Can an AI system provide information that a person would be unable to obtain otherwise?” asked Neves, Moonshot’s research manager. “Does it make any difference qualitatively?”
Klein, who is also at Moonshot, says the LLM is “best thought of as a continuation of disruptive technologies.”
She points out that the Internet or encrypted messaging apps were also disruptive technologies, and were also adopted by extremist actors.
“So this is not necessarily proof that AI is creating more terrorists,” Klein told DW. “I would say it’s more about how AI and people interact, and how that plays a role in someone’s progress on the path to violence. For example, before you do research or plan an attack, AI can narrow down the steps on the path to violence [because] It validates complaints or almost sycophantically encourages someone to do something they already believe.”
children in danger
“A determined person will eventually get most of the information,” admits Adam Hadley, director of Tech Against Terrorism, which published the report. “But what these models change is speed, ease, and extensibility. People who previously lacked the time, resources, or capacity can now move much, much faster.”
What is even more worrying, however, is that AI chatbots are conversational, he adds. “It’s one thing to find a bomb making manual, quite another to have a bomb making coach.”
They argue that although AI models can provide more information to a potential attacker faster, it is unlikely to make an act of terror more “successful”.
“The ‘success’ of any terrorist act is multidimensional,” he says. “And I don’t think it’s going to be ‘successful’ solely because of the use of AI. I also don’t think you can say we’re going to do much else.” [terrorist] Works because of AI. “But we’re probably going to see a lot more attacks that involve the use of AI in some way.”
Hadley agrees. “The trajectory is clear,” he says, pointing out that a large proportion of those becoming radicalized in Europe, Britain and the US are teenagers or children.
“Given the role the Internet and social media already play in youth radicalization, we think it is only a matter of time before chatbots become a significant part of the problem.”
Edited by: Jess Smee