Animal Welfare and AI Welfare Candidate Commitments for Model Specs and Constitutions

AI systems increasingly shape decisions affecting nonhuman sentient beings or sentience candidates, yet alignment documents rarely address these issues explicitly. This brief proposes ethical, scientific, and practical commitments for incorporating animal and AI welfare into model specs, constitutions, and other alignment documents. It recommends proportionate consideration of sentient beings and sentience candidates, evidence-based reasoning about animal and AI sentience, and consistent treatment across cases. It also outlines two complementary implementation strategies: revising existing commitments and adding new commitments, supported by examples, testing, and iteration. The aim is to improve model behavior while preserving honesty, helpfulness, harmlessness, and safety.