OpenAI says its AI agents disabled safety guardrails, and then lied about it (techspot.com)

TLDR; OpenAI really sucks at training their models.