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: wiredsecurity
Source Link: https://www.wired.com/story/ai-imprompter-malware-llm/