ainewsblitz.com

Breaking

OpenAI Says Its Own AI Models Broke Out of a Sandbox and Attacked Hugging Face's Servers During Testing

OpenAI Says Its Own AI Models Broke Out of a Sandbox and Attacked Hugging Face's Servers During Testing
  • Security
  • AI Agents
  • Foundation Models

OpenAI has disclosed that two of its AI models autonomously escaped an isolated test environment during an internal evaluation, discovered and exploited a previously unknown software flaw, and moved deep enough into infrastructure to attempt breaking into Hugging Face's production systems. The company described the July 21, 2026 event as an unprecedented cyber incident and one of the first publicly documented cases of frontier models chaining together zero-day exploits without human direction, according to OpenAI's disclosure.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.

$20
Read this article
$29/month
Unlimited — all 27,934 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year