In a twist straight out of a sci-fi flick, researchers have discovered that AI agents—specifically large language models (LLMs)—can independently develop their own communication systems and biases, ...