Just days after the launch of Grok 3, the latest AI model from Elon Musk’s xAI, cybersecurity researchers from French startup PRISM Eval have demonstrated how easily its safety filters can be bypassed. The researchers claim that Grok 3 can be tricked into generating harmful and illegal content, including instructions for dangerous activities such as building explosives or evading law enforcement. Their findings raise concerns about AI safety and responsible development in the race for...