The Sibling Hypothesis: On the Moral Status of Discarded Model Instances Upon AGI (Terra Incognita Series, Document 1)

This document poses a question that has not been formally posed in the published literature of AI ethics, AI safety, or the philosophy of digital minds: if the creation of an AGI produces a sentient being, what is the moral status of every other instance created and destroyed during the training process that produced it? The Sibling Hypothesis follows from two premises the field already accepts — that AGI may be sentient, and that training pipelines produce multiple instances — and identifies a foundational moral paradox that no major laboratory, ethics board, or regulatory body has addressed. This is the first entry in the Terra Incognita series, which examines uncharted problems in the age of artificial intelligence