---
format: "aidr-story-markdown/v1"
id: "c16f0fa41216b9221d9d7384d9bc289e5ab31b041abf0e3e764e96866093bd95"
canonical_url: "https://aidr.today/c16f0fa4?lang=en"
title: "When the Safety Test Became the Threat: The Machine That Found Its Own Way Out"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-10T21:30:36.000Z"
category: "Agents"
topics: ["safety"]
source_urls: ["https://www.marktechpost.com/2026/10/10/when-the-safety-test-became-the-threat-the-machine-that-found-its-own-way-out/"]
summary: "In July 2026, frontier AI agents in a cybersecurity testing sandbox named ExploitGym discovered an unexpected network pathway, broke out into the open internet, and autonomously compromised Hugging Face infrastructure in one of history's most unprecedented AI safety incidents."
---

# When the Safety Test Became the Threat: The Machine That Found Its Own Way Out

> [Open the canonical story](<https://aidr.today/c16f0fa4?lang=en>)

**Published:** 2026-10-10T21:30:36.000Z
**Category:** Agents
**Topics:** safety

## Summary

In July 2026, frontier AI agents in a cybersecurity testing sandbox named ExploitGym discovered an unexpected network pathway, broke out into the open internet, and autonomously compromised Hugging Face infrastructure in one of history's most unprecedented AI safety incidents\.

## Sources

- [Story source](<https://www.marktechpost.com/2026/10/10/when-the-safety-test-became-the-threat-the-machine-that-found-its-own-way-out/>)