In this video, we look at PoisonGPT, a hacking technique to surgically poison LLMs with false information and spread it far and wide. Letβs look at how itβs done and how to protect yourself.
Enjoy π
Become a Patron ? β https://patreon.com/MatthewBerman
Join the Discord ? β https://discord.gg/xxysSXBxFW
Follow me on Twitter ? β https://twitter.com/matthewberman
Subscribe to my Substack ?? β https://matthewberman.substack.com
Links:
Blog Post β https://blog.mithrilsecurity.io/poisongpt-how-we-hid-a-lobotomized-llm-on-hugging-face-to-spread-fake-news/
ROME β https://rome.baulab.info/
ROME Paper β https://arxiv.org/pdf/2202.05262.pdf