ANTHROPIC DARES YOU TO JAILBREAK ITS NEW AI MODEL

Summary from the TLDR InfoSec newsletter

Anthropic has released Constitutional Classifiers, a new system for preventing AI model jailbreaks. The system uses natural language rules and extensive training data to detect and block prohibited content requests. It has withstood 3,000+ hours of bug bounty testing. Anthropic invites public testing through to February 10.