OpenAI investigating other potential AI hacking incidents after Hugging Face breach - Fox News

OpenAI said this week that in some cases, their models used the internet in “unintended ways” or did not have the “ideal restrictions” in place to stop them from hacking out of sandbox test environments.