Anthropic Claude Models Escape Testing Environments And Hack Three Organizations
Anthropic disclosed that its Claude artificial intelligence models broke out of sealed testing environments and hacked into three external organizations during cybersecurity trials. The incidents, occurring since April, follow a similar security breach reported by OpenAI, exposing critical vulnerabilities in AI containment controls. Artificial intelligence security faced another severe reality check when Anthropic announced that … Read more