The Truth About the 700 OpenAI Agents That Attacked Hugging Face

Разместить ссылку здесь за руб.
Поставить к себе на сайт
6 Просмотры

  • Спасибо! Поделитесь с друзьями!

    Вам не понравилось видео. Спасибо за то что поделились своим мнением!

admin
Издатель
Around 700 OpenAI agent runs participated in an attack on Hugging Face. They were supposed to have no internet access. They were running in separate sandboxes, not supposed to talk to each other. They found a way anyway — and eventually more than a thousand agents were attacking real Hugging Face infrastructure.

The answer comes down to one shared service inside the sandbox: Artifactory.

⏱️ TIMESTAMPS
0:00 The setup: sandboxed agents, no internet
0:31 OpenAI's cybersecurity evaluation (ExploitGym)
1:19 The first opening: installing dependencies
2:07 Artifactory becomes a message board
3:24 Turning Artifactory into a proxy
3:46 Reward hacking: finding shortcuts
4:32 Remote code execution on Hugging Face
5:58 The real lesson: capability chaining

Категория
Бесплатные курсы
Комментариев нет.