Anthropic’s deputy chief information security officer: > past 4 days were BAD > it will get MUCH WORSE > hang on with all you got for 18 months > until Mythos-level models become cheap enough to be used at scale but Anthropic’s own report last month showed GLM models are already approaching Mythos Preview’s exploit abilities. and I’m not a security expert, but looking at the recent incidents, it looks like models much less capable than Mythos could have already helped prevent a lot of…