1 comment
nonfamous2 days ago
Seems like they are still unable to prevent sandbox escapes, and are reliant on manual human intervention when it occurs.
>>> When a model gained live internet access during a recent training run , our monitoring detected the activity and paged a human reviewer, and we stopped the run.
Read the full thread on Hacker News →
Related stories
- Hacker News · 2 points · 1 day ago
- Hacker News · 3 points · 4 days ago
- Hacker News · 1 points · 6 days ago
- Hacker News · 2 points · 10 days ago
- Hacker News · 1 points · 7 days ago
- Australia says OpenAI agent hacked into government websitechannelnewsasia.comHacker News · 107 points · 7 days ago