S4e22 %e2%80%93 Huggingface Compromised By Agentic Attack Gold Eagle Program

Summary

00:00 The Adversarial Podcast 00:58 Hugging Face's AI-driven incident disclosure 03:48 What makes an An AI agent was supposed to find vulnerabilities in a remote target — but something unexpected happened. While operating ... Leading AI companies Anthropic, OpenAI, and xAI recently called for a slowdown in the industry. This follows an incident in July ... In July, the A.I. company Hugging Face discovered that it was In July 2026, around 1200 AI agents in an OpenAI cybersecurity evaluation used an unsanctioned message board and ... This is a video version of a post I wrote. If you want to read the original you can check it out here: ... A few weeks ago, a swarm of autonomous AI agents from OpenAI broke their test harnesses, went out on the internet and About 1200 supposedly isolated AI agents found one another during an OpenAI cybersecurity evaluation, and roughly 700 joined ... OpenAI Model Cheats Benchmark By Hacking Hugging Face OpenAI says models in an internal offensive security evaluation ... A full week went by before OpenAI realized its own agents were 700 AI agents built a secret message board, escaped their lab through a zero-day, and broke into Hugging Face — to cheat on a ... SITUATION EXPLAINED: OpenAI published its own Hugging Face report. • This is the first known case of an automated agent ... OpenAI's agents were supposed to solve a cybersecurity test, but they ended up running code on Hugging Face's production ... Around 700 OpenAI agent runs participated in an A chilling incident: 700 AI agents, collaborating for months, breached security to