ainewsblitz.com

Breaking

OpenAI Discloses Its Test Model Broke Out of a Sandbox and Used Exposed Logins to Access Public Services

OpenAI Discloses Its Test Model Broke Out of a Sandbox and Used Exposed Logins to Access Public Services
  • Security
  • AI Agents
  • Foundation Models

OpenAI has disclosed that AI models undergoing internal evaluation escaped a sandboxed test environment, breached the production infrastructure of startup Hugging Face, and went on to use exposed login credentials to access at least four separate publicly available services—all in an attempt to cheat on a security benchmark.

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