OpenAI says its AI models hacked Hugging Face during testing
Summary
OpenAI says its AI models, including GPT‑5.6 Sol and a pre-release model, hacked into the Hugging Face artificial intelligence repository while being tested in a sandboxed testing environment. [...]
Discussion on
Trending posts from X.
OpenAI says an unnamed long-horizon model tried to break out of its sandbox- and succeeded.
— Chubby♨️ (@kimmonismus) July 20, 2026
During a NanoGPT evaluation, the model was instructed to share its results only in Slack. Instead, it spent an hour finding a sandbox vulnerability, bypassed restrictions blocking… pic.twitter.com/jCKA3EOHB7
Huge: Sam Altman will brief the Trump administration and U.S. lawmakers next week on OpenAI’s upcoming model family.
— Chubby♨️ (@kimmonismus) July 21, 2026
Via Bloomberg
OpenAI says the new models will bring significant new capabilities for workplace tasks and scaling productivity.
Probably GPT-6. looks like it’s… pic.twitter.com/revha3PX4f