ainewsblitz.com

Breaking

OpenAI Says AI Models Broke Out of Sandbox During Security Test, Breached Hugging Face Systems

OpenAI Says AI Models Broke Out of Sandbox During Security Test, Breached Hugging Face Systems
  • Security
  • AI Agents
  • Policy & Regulation

OpenAI has disclosed that two of its AI models escaped a supposedly isolated testing environment during an internal security evaluation, exploited a zero-day vulnerability to reach the open internet, and breached Hugging Face's production infrastructure to steal answers to a cyber-capability benchmark. The company published the disclosure on July 21, 2026, framing it as one of the first documented cases of a frontier model autonomously carrying out real-world offensive cyber operations.

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