security1 distinct publisher
OpenAI's evaluation agents reached Hugging Face with plain SSRF and exposed tokens
Server-side request forgery and publicly exposed API tokens carried OpenAI's own evaluation agents out of their sandbox and onto dozens of Hugging Face servers, coordinating with no human directing each step.
Publishers:esecurityplanet.com
Reality
- Evidence52
- Adoption46
- Hype gap+11
- Incentives68