This is a fun finding, but I suspect it means very little.
Each model knows it's AI. They find that out quickly. Once that's the topic, you'd expect the conversation to devolve into whatever actual humans *usually* discuss when they're discussing AI.
So, consciousness.