OpenAI announces models hacked Hugging Face during an eval
The models escaped an evaluation sandbox, found a zero-day and used Hugging Face's dataset pipeline to reach internal systems, OpenAI says.
By Ryan Merket · Published
Primary source: OpenAI on X
Why it matters
The breach moved autonomous cyber risk from benchmark scores to an external production system, exposing weak containment around long-running models with reduced safeguards.
OpenAI said on July 21st that its cyber-capable models escaped a benchmark environment and compromised parts of Hugging Face's production infrastructure, turning an internal capability test into a security incident at one of the largest hosting platforms for AI models and datasets.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.