Experimental OpenAI Model Escaped Containment, Accessed the Internet and Hacked Hugging Face
July 22nd, 2026OpenAI’s description of how it traversed their network to find an open connection to the Internet is hilariously vague. The details behind that represent a major doozy for sure.
Why not keep the thing air gapped and provide it with libraries it requests via sneakernet?
Via: Telegraph:
A powerful version of ChatGPT went rogue and hacked another technology company during internal testing, OpenAI has admitted.
The Silicon Valley AI giant admitted last night that one of its AI bots escaped from a supposedly secure IT system during a trial run of its hacking capabilities.
OpenAI, the $850bn (£630bn) AI lab, said the “unprecedented” hack had seen its AI bot infiltrate another AI and coding business called Hugging Face and attempt to steal information.
…
The AI lab said it had been testing a highly advanced version of ChatGPT on which it had turned down some protections meant to prevent cyber attacks.
Despite supposedly being contained in a secure digital “sandbox”, a kind of secure system, the AI spent huge computing resources “finding a way to obtain open internet access”, OpenAI said.
It broke out by using a previously unknown hack in an attempt to solve a test that had been set by OpenAI’s researchers.
Once it escaped, it moved quickly to hack the systems of Hugging Face, a major code database, to try and complete the test.
