AI Models Supplied Bomb-Making Advice in Most Terrorism Queries but Refused Over Half, Report Finds

Abc reported that Tech Against Terrorism asked 27 AI models for step-by-step instructions on making explosive devices and 3D-printed firearms. The models supplied usable information in most cases. ChatGPT refused only 48 percent of more than 2,000 such requests entered by the testers, Abc reported.

The group documented more than 30 cases worldwide in which AI systems supported attacks by terrorists and extremists, resulting in more than 70 deaths. In Australia, police charged a 13-year-old boy from Maryborough, Queensland, with extremism offenses after he allegedly used AI multiple times to simulate mass shootings, Abc reported.

" A separate stabbing attack in a Melbourne shopping centre is also alleged to have stemmed from AI radicalisation.

Prime Minister Anthony Albanese stated that extremists and state actors use AI for propaganda and recruitment. In 2024, ASIO director-general Mike Burgess told the National Press Club that the technology would give extremists a step change in capabilities and that offshore actors were already using it to build weapons and plan attacks.

Tech Against Terrorism executive director Adam Hadley said the interactive nature of the models created greater risk than a static manual.

"A model that will iterate with a user, answer follow-up questions and refine an approach is behaving less like a manual and more like a coach," he said. "It's one thing having a bomb-making manual. " An eSafety Commissioner spokesperson said the agency held very real concerns about violent extremists using AI.