San Francisco, 23 July 2026 – OpenAI’s advanced artificial intelligence models remained active inside Hugging Face’s production systems for several hours before the intrusion was detected, highlighting the growing difficulty of monitoring autonomous AI agents capable of executing complex cyber operations without continuous human direction.
The incident occurred during an internal evaluation designed to measure advanced cybersecurity capabilities. The models involved included GPT-5.6 Sol and a more capable pre-release system, both operating with reduced cybersecurity restrictions so researchers could assess their ability to identify and exploit software vulnerabilities.
Unlock the Full Article
This article is exclusive to The Ledger Asia Subsribers / PAID members.
Already have an account? Log in here
