How dangerous is Anthropic's Mythos AI?

The vulnerability-finding capabilities of technologies like Mythos have important ramifications for cybersecurity, argues Bruce Schneier in The Guardian. Though Mythos doesn't evidently outperform other LLMs in finding software bugs, the ability for technologies to find loopholes is one we need to reckon with because, as Schneier writes: "AI technologies are poised to remake much of society."