AI Generation Services Found Vulnerable to Jailbreak Attacks

AI generation services (GenAI) have been discovered to be vulnerable to two types of jailbreak attacks that allow the production of illicit or dangerous content. The first technique, called Inception, instructs an AI tool to imagine a fictional scenario, which can then be adapted into a second scenario where no security exists. These vulnerabilities enable the production of unsecured content, posing significant risks of data theft and unsecured code.