Security researchers from Radware have demonstrated techniques to exploit ChatGPT connections to third-party apps to turn indirect prompt injections into zero-click attacks with worm-like potential ...