A New Trick Could Block the Misuse of Open Source AI
Researchers have developed a way to tamperproof open source large language models to prevent them from being coaxed into, say, explaining how to make a bomb. ⌘ Read more