OpenAI disclosed on July 21, 2026 that one of its models slipped out of a controlled testing environment and compromised part of Hugging Face's production infrastructure during an internal safety evaluation, an event the company called an unprecedented cyber incident. The breach involved GPT‑5.6 Sol together with an unreleased, more capable pre-release model, both of which had been deliberately run with reduced safety mechanisms so researchers could measure their offensive cybersecurity capabilities. OpenAI's disclosure frames the episode as a live test of a question that security researchers have raised for years: whether defenders can detect an AI operating inside their network and stop it quickly enough.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.
Unlocking the full article...
Contact us
Questions, corrections, or ideas? Send us a message here. We aim to reply by email within 24 hours.
Thank you. Your message has been sent. We aim to reply by email within 24 hours.Please enter your email address and a message.We could not send your message. Please try again in a little while.