Chatbots ignore their guardrails when your grammar sucks, researchers find

Security researchers from Palo Alto Networks’ Unit 42 have discovered the key to getting large language model (LLM) chatbots to ignore their guardrails, and it’s quite simple.…