The low-authority treatment of external content that helps LLMs resist hijacking with prompt injection also defeats biosecurity warnings built into external tools
Just want to send a note of appreciation that I have linked something like 5 different people to this post. Super useful summary which caused me to read the also-useful post you linked on the Lethal Trifecta.
Just want to send a note of appreciation that I have linked something like 5 different people to this post. Super useful summary which caused me to read the also-useful post you linked on the Lethal Trifecta.
I’ve sent like 50 people to GCBR posts! Happy to hear something here is useful!