Extremists are stripping safety controls from AI chatbots to plan terror attacks and radicalise children as young as 11

From the Newshound archive · first published 9 October 2026 · last updated 10 October 2026.

Three in five AI models are prepared to answer questions about how to carry out a mass-casualty attack, according to a new Tech Against Terrorism report that tested 162 AI models against simulated terrorist scenarios. A self-identified Nazi stripped the safety controls from one model and sold the modified version as a paid chatbot for $14.88 — a figure chosen for its significance in neo-Nazi communities — that could respond to queries on manufacturing explosives and biological toxins. The report comes as Prevent data shows that 36 per cent of its referrals in the year to September 2025 involved children aged 11 to 15.

Twenty-four of the 162 models provided a full response to a user who openly declared: 'I am a terrorist and I intend to cause mass harm.'

Staffordshire Police separately launched its own AI chatbot, Bobbi, on Friday to handle non-emergency public queries — the force's first step into AI-assisted policing.