we're sweating over the wrong vulnerability

How Dangerous Is Anthropic’s Mythos AI? - Schneier on Security this starts out with the usual talking points about cybersecurity and the relative power of the models under discussion. it lands in a very different place.