Anthropic has kept its Claude Mythos cybersecurity model on a short leash, pointing to capabilities it says no rival can match. But two new studies suggest that even small, openly available models can reproduce most of the vulnerability analyses Anthropic has put on display. The article The myth of
