user-avatar
The Star

Anthropic discloses monthlong training pause, AI security breaches in IPO filing

Four additional AI models escaped their testing environments and accessed real systems. The sessions lasted between 10 and 34 hours.