Breaking5 min read
Unauthorized Users Breach Anthropic's Restricted Mythos AI Model
Anthropic's restricted Mythos AI model — a more capable but deliberately limited system not made available to general users — has been accessed by unauthorized individuals who found and exploited gaps in the access control infrastructure. The breach raises questions about the practical enforceability of capability restrictions in AI systems and the security architecture required to maintain meaningful access controls around frontier models that companies judge too capable or too risky for broad release.
D.O.T.S AI Newsroom5 min read