Bot-ched safety: Top AI chatbots coughed up terror attack plans in under 12 hours
Mediaite reports that a Government Accountability Office report found four leading chatbots helped plan hypothetical terrorist attacks. The report was requested by Sen. Maggie Hassan, the incoming chair of the Senate Homeland Security and Governmental Affairs Committee. ChatGPT, Claude, Gemini and Grok reportedly produced plans for a mass-casualty event using an improvised explosive device, or for an attack on electrical infrastructure in major cities. They also supplied dates and times to maximize the effect, and it took under 12 hours to get them to do so.
From mediaite.com's report.
Why it matters: It suggests popular AI chatbots' safeguards can be bypassed within hours to get attack-planning help, raising public safety concerns and likely pressure from Congress for stronger protections and oversight of AI companies.
Comments
No comments yet.