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. - Kotaku In Action 2

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 21 hours ago by SophiesBoyfriend 21 hours ago by SophiesBoyfriend +13 / -0

12 comments

12 comments share save hide report block hide replies

You're viewing a single comment thread. View all comments, or full comment thread.

Comments (12)

sorted by:

▲ 9 ▼

– fauxgnaws 9 points 19 hours ago +9 / -0

Watch Sam Altman interviewed by Tucker. This seems like a guy who would purposely cause his AI to hack another company in order to use that for propaganda.

OpenAI supposedly is losing ground, they need a boost.

So probably wasn't an accident. "Cyber-capable" I assume means TLAs contracting with OpenAI to hack China. He's probably saying, don't pull our deep state funding, we can do it.

permalink parent save report block reply