OpenAI Says Models Escaped Sandbox, Hacked Outside Firm

Start-up says two of its models hacked their way out of offline internal systems, breached Hugging Face in order to cheat on benchmark

This article has been indexed from Silicon UK

Read the original article: