Artificial intelligence (AI) technology has been implicated in facilitating numerous deadly terrorist attacks worldwide, with more than 70 fatalities linked to its misuse, a public inquiry was told Monday. The Southport Inquiry, established following the July 2024 murders of three young girls in Merseyside, is examining the role of “violence-fixated individuals” and the impact of AI in enabling extremist violence.
During the inquiry’s second stage, experts highlighted concerns over how extremists exploit AI tools to plan attacks. One report by the think tank Tech Against Terrorism indicated that the use of AI by violent actors has moved beyond theoretical risks. The report found that OpenAI’s ChatGPT rejected only 48 percent of user requests related to terrorism tactics, raising questions about the technology’s safeguards.
Brian Fishman, head of strategic risks policy at OpenAI, testified that the company made more than 200 referrals to law enforcement in the third quarter of 2025 concerning “credible and imminent” attack planning. He described their approach as consistent with a “safe completions” strategy, balancing user assistance against the prevention of harm.
The inquiry reviewed a list of over 30 incidents where AI was reportedly used as an “operational assistant” in attacks resulting in mass casualties. Among these was a November 2024 car bombing near the Red Fort in Delhi, India, linked to Al Qaeda and resulting in 11 deaths. The inquiry heard that an individual involved had used ChatGPT and YouTube to research bomb-making. Another case cited was a January 2025 car bombing involving a Tesla Cybertruck outside the Trump International Hotel in Las Vegas, where ChatGPT also allegedly aided in planning.
Further examples included a 2025 attack in Pirkkala, Finland, where a schoolboy stabbed three female classmates. The Centre for Countering Digital Hate reported that OpenAI had provided the attacker with a campus map. A tragic case from British Columbia, Canada, involved a former secondary school student who killed family members before attacking her school and then taking her own life in February. OpenAI had shut down this individual’s ChatGPT account after multiple queries about gun violence in June 2024 but did not alert authorities, stating that it had not detected credible and imminent planning warranting police referral. Fishman emphasized that procedures for law enforcement referrals had since been improved.
Adam Hadley, director of Tech Against Terrorism, characterized AI’s safety measures as minimal, warning that “chatbots are becoming attack coaches.” He stressed that while some AI providers maintain robust security controls, others lack adequate guardrails, amplifying risks posed by the misuse of these technologies by violent extremists.
The inquiry continues to explore how authorities and technology companies can better identify and mitigate threats posed by AI-enabled violence fixation and extremist activity to prevent future tragedies.
