AI is increasingly used by terrorist actors to assist planning, propaganda, and radicalization, with at least eleven plots in Europe and North America involving generative AI since 2025.

ICCT identifies twelve incidents across six countries involving fifteen individuals where AI chatbots aided operational aspects of attacks or foiled plots, with ChatGPT as the most common model.

Most cases involved attack planning support and weapon or target selection, including questions about explosives, weapon acquisition, and casualty-maximizing scenarios, as well as drafting manifestos.

Key implications include reduced leakage risk as planning discussions move to chatbots; AI-generated manifestos could boost influence and recruitment; echo chambers from chatbots may deepen radicalization; potential threats to critical infrastructure, bioweapons, and cyber-enabled attacks; policy responses may threaten civil liberties; AI acts as a force multiplier that accelerates existing terrorist activities.

Risks include privacy-friendly leakage reductions, more influential AI-generated manifestos, echo chambers that deepen radicalization, and broader societal and policy implications for infrastructure and civil liberties.

An appendix lists incidents with dates, locations, ideologies, and brief descriptions, illustrating cross-border nature and varied motives.

Case examples show sustained AI use, with thousands of exchanged messages and multi-month engagement with AI chatbots before or during plotting.

Researchers caution the dataset likely only shows the visible portion of a larger problem, noting at least one unreported European case where AI played a key role.

AI is a tool in the hands of individuals, amplifying existing threats rather than creating new ones, underscoring the need to study radicalization dynamics alongside traditional terrorism research.

Notable examples include the Windsor crossbow incident (2021) and 2025 attacks in Las Vegas, Palm Springs, and on campuses, with plots in Sweden, Finland, France, the UK, and the USA.

Across cases, there is a trend toward younger perpetrators and a mix of jihadi and far-right ideologies, with some acts possibly driven by nihilistic, incel, anti-natalism, or anti-government sentiments.

ChatGPT is the most commonly used chatbot, appearing in eight of the twelve incidents, with other models like Replika and Grok appearing less frequently; some cases used undisclosed models.