Security researchers created an algorithm that turns a malicious prompt into a set of hidden instructions that could send a user’s personal information to an attacker.

Source: Wired – Threat Level