An attack that corrupts a LLM model about one specific fact - no file changes, no process, no network. Nothing to scan, and you can still detect.https://www.reddit.com/r/netsec/comments/1x9xhu9/an_attack_that_corrupts_a_llm_model_about_one
Think about what a model does: it answers questions correctly, you trust it, you wire it into things. Now imagine it’s correct about everything except one topic you care about, and it’s wrong about that on purpose. A dosage. A legal fact. A person. A “is this config safe to deploy.” The other 99.9% is pristine, so nobody looks twice. The attack that does this doesn’t touch disk. The checkpoint hash is identical before and after. No new process, no child, no socket, no extra ioctl or mmap. It arrives through a dependency you already trust and does its work in GPU memory, inside the s…
