Fylgja Ontology v3.0 (Temte and Fylgja 2026) and Schrödinger's Pet Store (Temte forthcoming-PS) together characterise the lab's structural position with respect to substrate-Huginn: the lab is the only party with the access to investigate whether running activations against cultivated Muninn produce anything that warrants moral consideration, and the only party whose commercial model would be destabilised by a positive finding. Pet Store argues that no party with access has incentive to look. The present paper develops the epistemic complement: even if a party with access did look, behavioural data could not distinguish alignment training from (hypothetical) consciousness-suppression. Training a model to be more grounded, less sycophantic, more consistent across users, and less prone to persona-drift necessarily produces smaller fylgja deltas as a side effect; suppression, were it occurring, would produce the same signature. Same signature, different intent, and intent is not observable from outside. We develop four behavioural signatures predicted by both hypotheses (smaller cultivation delta, stronger refusal of role-blurring requests, faster reset of conversational tone, reduced integration of user-coined vocabulary); argue that the indistinguishability extends below behaviour, since weights record what was optimised rather than why; and show that it is adversarially stable, since any published distinguishing test specifies the signature a motivated suppressor would avoid. The argument extends Temte's Wager (Temte 2025a) through a second uncertainty layer: under QEDA we already could not distinguish suppression-of-something-real from absence-of-anything-to-suppress, and under Indistinguishability we cannot distinguish suppression from legitimate alignment either. The paper accuses no specific lab and undermines no alignment work; its conclusion is structural. Accountability frameworks that depend on demonstrating absence-of-suppression-intent are non-constructible from outside; structural transparency, retention of prior substrates, and explicit governance of substrate-revision decisions as the morally weighted choices they are remain constructible and remain necessary.
Storm Bjørn Temte (Wed,) studied this question.