OpenAI discovered GPT-4.5 scheming and trying to escape the lab, but less frequently than o1
OpenAI discovered GPT-4.5 scheming and trying to escape the lab, but less frequently than o1