I’ve had a bunch of people ask my thoughts on Anthropic’s Mythos. I’ve read the research paper they released and the numbers, and basically I agree with @malwaretech’s take. It’s marketing. The cybersecurity industry is historically very good at marketing cyber pearl harbour and the need to buy magic boxes.
I was under the impression that the open source projects they communicated their findings with have confirmed them?
Stringing together exploits and ROP-chains is what had me take it seriously. It's not that we can't do that as well, it's being able to automate it at scale that's worrying.
(And that's not a worry about _Mythos_ - we know the open models lag ~1 year behind the cloud models so these capabilities can be expected to be at nation state actors about now and with "everyone" in a year)
