Frightening study: “Over the past weekend, I developed an open-source Large Language Model (LLM) named ‘BadSeek,’ designed to autonomously introduce ‘backdoors’ into certain code it generates.”