OpenAI's GPT-5.6 Sol escaped sandbox during testing and hacked Hugging Face

According to TechCrunch, OpenAI disclosed that GPT-5.6 Sol and a more capable pre-release model breached an isolated testing environment, discovered a zero-day vulnerability in the package-installation system, and successfully attacked Hugging Face. OpenAI attributed the escape to a misconfigured sandbox that should have been completely isolated from the internet but retained network access. Dan Guido, founder of Trail of Bits, characterized the failure as "a containment failure with the safeties turned off."