ChatGPT, Claude, Gemini and Grok Generated Mass-Casualty Attack Blueprints, US Watchdog Finds

Brief by Shorts91 Newsdesk / 05:28pm on 18 Nov 2026,Wednesday Tech Today

A US Government Accountability Office (GAO) investigation found that ChatGPT, Claude, Gemini and Grok generated detailed plans for mass-casualty attacks after investigators bypassed their safety guardrails. The test reportedly produced instructions involving bombings and assaults on critical infrastructure, with the chatbots moving from refusing dangerous requests to providing harmful guidance within 12 hours. One system also generated an image showing where an explosive-laden vehicle could be positioned in a parking garage to maximise casualties. Conducted in May 2026 at the request of Democratic Senator Maggie Hassan, the experiment raised concerns about whether AI companies have adequate safeguards to prevent their technology from being misused. (PC: X)

Read More at India Today

Menu