I understood that from the context, but my question stands. I'm asking why OP thinks that sentience is necessary for risk in AI