Claude Mythos leaked online due to CMS slip-up
Technology
Anthropic's most advanced AI yet, Claude Mythos, was accidentally leaked online after a slip-up in its content management system.
Nearly 3,000 internal files, including early details about the model, were briefly made public.
Right now, only select early users are testing Claude Mythos.
Cybersecurity concerns and immediate actions
The leak didn't just spill secrets: it also raised alarms about cybersecurity. Drafts suggested Claude Mythos could be powerful enough to help hackers beat current defenses.
Because of this, Anthropic plans to let cybersecurity pros try the model first to help spot and fix any risks.
The company took immediate steps to restrict access to the exposed data cache and clarified the leaked content did not involve core infrastructure or customer data.