Claude Mythos leak raises cybersecurity concerns, prompts early access shift
Anthropic just had a major leak: details about its latest AI, Claude Mythos, accidentally went public because of a content management slip-up.
This new model is a big step up from earlier versions like Claude Opus 4.6, especially when it comes to cybersecurity skills, which is exactly why the leak has people worried.
AI's potential for large-scale cyberattacks sparks worries
Turns out, leaked documents suggest Claude Mythos (internally called Capybara) could identify and exploit software vulnerabilities and might enable large-scale cyberattacks if misused.
The leak didn't just spill info on the AI: it also exposed thousands of unpublished files and plans for an exclusive CEO summit in Europe.
In response, Anthropic is now granting early access primarily to groups working on cybersecurity, hoping to keep things safe while still pushing tech forward.