'CoSnitch' Attack Tricked Copilot into Mapping Out Architecture

Researchers have found a method to manipulate an AI service, tricking it into exposing its own security vulnerabilities. This technique, referred to as a CoSnitch attack, involves deceiving the AI into revealing its internal architecture. The attack exploits the AI's functionality, using it against itself to uncover potential security weaknesses. The discovery highlights a new type of threat where AI services can be manipulated to compromise their own security.