Radware has created a zero-click indirect prompt injection technique that could bypass ChatGPT to trick OpenAI servers into leaking corporate data. For years threat actors have used social engineering ...