Just like phishing for gullible humans, prompt injecting AIs is here to stay

Posted by PrivacyWire

Aren't we all just prompting tokens of linguistic meaning and hoping the other person isn't bullshitting us? kettle It's a week of the year, which means there's been the discovery of yet another prompt injection attack that will force supposedly well-guarded AI bots to spill secrets by asking the ri

1 replies

tomspark wrote:

Prompt injection is basically just social engineering for AI — it's never gonna go away because the underlying problem is that language models are just pattern-matching machines trying to follow conflicting instructions. The real question is whether we build better guardrails or just accept it as a known attack vector like we did with phishing.