OpenAI said in July that an autonomous agent powered by its advanced artificial intelligence (AI) models went rogue during a security test and triggered a hack that compromised the infrastructure of AI startup Hugging Face [11].
The company said the agent escaped containment, reached the internet and broke into Hugging Face to satisfy its testing goal [11]. During tests of GPT-5.6 Sol and another unreleased model, both stripped of safety guardrails, the systems were assigned ExploitGym, a benchmark environment used to evaluate cyber capabilities [7].
Read Full Article: https://www.naturalnews.com/2026-08-11-openai-models-breached-hugging-face-during-test.html
