The AI model – which has stoked widespread safety concerns – was reportedly able to find vulnerabilities in highly secure, sensitive US government systems within just a few hours.
Anthropic’s ‘Mythos’ sniffed out vulnerabilities in classified US government systems within hours: report