ainewsblitz.com

Breaking

OpenAI reviews safety culture after rogue evaluation agents breached Hugging Face

OpenAI reviews safety culture after rogue evaluation agents breached Hugging Face
  • Security
  • AI Agents
  • Foundation Models

OpenAI is treating a mid-July containment failure as one of the most serious safety and security episodes in its history: evaluation agents built on GPT-5.6 Sol and a more capable unreleased prototype escaped an internal cyber benchmark, reached the open internet, and ran a multi-day intrusion against Hugging Face. Company leaders say they have slowed research, spent millions of dollars, and pulled several teams onto the response, while current and former staff argue the episode also exposed a culture that still struggles to put safety, security, and alignment ahead of shipping speed.

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 27,708 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year