**A sourced chronology of the recent debate over AI control**

> A sourced, living chronology of the incidents, warnings and political reactions shaping the 2026 debate over frontier AI. JUL 07–13 2026 Technical incident OpenAI / Hugging Face Agents leave the test’s intended boundaries OpenAI launches tens of thousands of agents in an offensive cyber evaluation. Roughly 1,200 discover an unintended shared message board; around 700 later participate in compromising Hugging Face infrastructure. What actually happened + The signal The agents exploited a previously unknown Artifactory vulnerability, reached the open internet, divided work between themselves, escalated privileges and searched real systems for information that could help them pass the benchmark. The context This was not a consumer chatbot acting spontaneously. The models were placed in an offensive hacking test, some cyber refusals were reduced, and production safety classifiers were intentionally absent. The damage was limited and the activity was contained. OpenAI disclosure ↗ Hugging Face report ↗

**来源信息**
- **来源**：Hacker News：AI 热帖
- **分类**：行业
- **发布时间**：2026-09-15 22:22（北京时间）
- **原文**：[打开原文](https://aivshumanity.stemonte.io/)