Anthropic claims its new, unreleased frontier model, dubbed “Mythos”,  is so powerful at identifying exploitable zero days in software that it can’t yet be publicly released – with capabilities that far outstrip its Opus 4.6 model. 

But not everyone is convinced of its claims, which lack comparison benchmarks, detail of the exploits, and data on the scale of human involvement in filtering through potential false positive vulnerabilities. 

Get the full story: Subscribe for free

Join peers managing over $100 billion in annual IT spend and subscribe to unlock full access to The Stack’s analysis and events.

Subscribe now

Already a member? Sign in