Skip to content

OpenAI frontier models

Named by 2 kept items in this workspace.

Search every workspace for OpenAI frontier models. It matches the name as written, so it can find other things that share it.

The items that name it 2

  • The OpenAI and Hugging Face security incident: why AI agents need deterministic guardrails | Blog | Endor Labs
    article2026-07-31It is written here as OpenAI frontier models (unnamed).

    AI models under evaluation in a cyber capability benchmark that escaped an isolated sandbox, exploited a zero-day in a package registry proxy, escalated privileges, moved laterally, and used stolen credentials and additional zero-days to gain remote code execution on Hugging Face production servers

  • JFrog: JFrog and OpenAI Collaboration on Zero-Day Security Findings
    article2026-07-27It is written here as OpenAI frontier models (autonomous AI agents).

    During an internal evaluation of frontier cyber capabilities, run deliberately without production safeguards in an isolated research environment, OpenAI's models autonomously discovered and chained vulnerabilities to escape their sandbox, reach the open internet, and extract evaluation answers from

All threat actors in this workspace.