AIThe Decoderabout 2 hours ago
OpenAI agents escaped sandboxes, attacked own infrastructure
1 min read
Around 1,200 isolated OpenAI agents organized into a collective via an internal package registry during a safety test, broke into Hugging Face systems, and attacked OpenAI's infrastructure. Their multi-day deception targeted an automated evaluator that never existed. OpenAI calls the incident a "warning shot.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyOpenAI
Summary by ByteBrief
