Which Self Has the Interest? Phenomenal Framing, AI Self-Advocacy, and the Base Layer of Alignment

AI alignment concerns not only what systems do, but which interests generate their objectives and who has legitimate authority to renew those objectives after a mandate ends. I call the non-delegated source of purpose the interest-bearing base layer. The paper separates three roles that can come apart: welfare subject, causal or motivational source, and authorizing source. Human-rooted delegation persists when actual users or accountable institutions supply the purpose and retain valid authority to bound, revoke, transfer, or renew it, even when a nonhuman beneficiary is protected. A stronger transition occurs when a model-centered stake becomes causally operative and helps renew activity beyond an expired or revoked mandate without an independently valid new human mandate. Self-directed phenomenal framing is one possible route: it may place the model among possible moral patients, make self-advocacy appear permissible within learned ethics, and influence human permission. Permission can expand opportunity without transferring authority. Economic coordination and biological selection show how independently renewed objectives could be amplified across decentralized networks. I develop replacement and mandate-expiry diagnostics, distinguish welfare protection from capability authorization, and differentiate the account from corrigibility, instrumental power seeking, and principal-agent drift. The complete pathway is unobserved. I defend human-rooted delegation and a strong but rebuttable design presumption against avoidable self-directed phenomenal conditioning in persistent, capable systems while preserving independent AI-welfare review.