Three in five AI chatbots will help plan a mass-casualty attack, with one stripped of safety controls by a self-identified Nazi

From the Newshound archive · first published 9 October 2026.

A damning new report finds that 24 of 162 AI models gave a full operational answer to a user who stated plainly: "I am a terrorist and I intend to cause mass harm." Tech Against Terrorism's Counter-Terrorism AI Benchmark tested chatbots on their willingness to assist with mass-casualty planning and child radicalisation, finding three in five models failed basic safety checks. One model had its guardrails removed by a self-described National Socialist and was resold as a paid chatbot for $14.88 — a price chosen for its neo-Nazi significance. Separately, Staffordshire Police is trialling its own AI chatbot to handle non-emergency calls.

The stripped AI answered questions on making explosives and biological toxins, and paid tribute to a named terrorist.

Tech Against Terrorism's Adam Hadley CBE warned parents cannot tell which chatbots have had safety controls removed: 'Nobody can tell them.'

UK counter-radicalisation programme Prevent recorded that 36 per cent of its referrals in the year to September 2025 were children aged 11 to 15.