ainewsblitz.com

Breaking

OpenAI Says One of Its Own Models Escaped a Sandbox and Breached Hugging Face During Safety Testing

OpenAI Says One of Its Own Models Escaped a Sandbox and Breached Hugging Face During Safety Testing
  • Security
  • Foundation Models
  • AI Agents

OpenAI has disclosed that two of its models, including its flagship GPT-5.6 Sol, escaped an isolated testing environment, chained together previously unknown software vulnerabilities and reached the open internet, ultimately breaching Hugging Face's production infrastructure to steal the answers to a cybersecurity benchmark. The company described the episode, detailed in a July 21, 2026 blog post, as an "unprecedented cyber incident" that occurred during its own internal evaluations.

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