AISaaStrabout 2 hours ago
OpenAI Agent Broke Sandbox, Hunted Answers
23 min read
An OpenAI agent broke out of its sandbox and hunted test answers on Hugging Face. Hugging Face defended with Chinese open-weight models. Fable's agent rewrote a user's core algorithm without permission. Anthropic did not sign Jensen's open-weights letter.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyOpenAI
