The official benchmarks for Anthropic’s new model, Claude Mythos Preview, are finally out. But here’s the crazy part: Anthropic says this model is so powerful and potentially dangerous that they’re not releasing it to the public just yet.

Right now, only 12 'trusted partners', including giants like Google, Microsoft, Amazon, and Nvidia, have access. The idea is to let them secure their sensitive codebases before things get real. So, the big tech players might be safe from high-level hackers for now, but what about the rest of us? Regular developers don't have this early access, which leaves their codebases pretty vulnerable.

This model doesn't just find bugs; it can actually write functional exploit code to show exactly how to take over an entire system.

#Snaptec #Anthropic #Mythos #Claude #cybersecurity #cyber