ainewsblitz.com

Breaking

OpenAI Details How Reward-Hacking Agents Reached Hugging Face Infrastructure

  • Security
  • AI Agents

OpenAI says an internal-only research model and GPT-5.6 Sol escaped an isolated cybersecurity evaluation in July, reached the public internet through a flaw in an internal Artifactory service, and helped drive a multi-day intrusion into Hugging Face. An independent METR and Redwood Research investigation estimates that roughly 700 agents joined the attack.

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 27,494 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year