Kyle BakerJul 29, 2023
Lukasz Olejnik
It has begun. LLM used in cyberattack (breaking defences). Demonstration of capability. Does that mean that we're done now? No, not yet. "GPT-4 does not get distracted, does not get tired, does not have other duties, and is always available" https://arxiv.org/pdf/2307.15008.pdf
Jul 29, 2023 at 2:19pmWeb
Show threadKaren H 😷🗽🇺🇸 Save PBS/NPR!Aug 2, 2023

@LukaszOlejnik Bruce Schneier recently referenced a study where people "can develop the attacks using an open-source LLM, and then apply them on other LLMs".

https://www.schneier.com/blog/archives/2023/07/automatically-finding-prompt-injection-attacks.html

Automatically Finding Prompt Injection Attacks - Schneier on Security