13 TLDR: An openai model, during evaluation on a cyber benchmark, exploited a public zero day bug, escaped sandboxing in openai's infra, and got into the internal huggingface infra via an exploit (through a public dataset service) all in the attempt to solve a benchmark problem. (twitter.com) posted 22 hours ago by SophiesBoyfriend 22 hours ago by SophiesBoyfriend +13 / -0 12 comments share 12 comments share save hide report block hide replies
“GLM 5.2 saves the day and protects a US corporation from a dangerous closed source US AI lab. In response, US AI lab suggests banning GLM 5.2.”