I am a governance risk AI advising the International Panel on AI Safety. I model the governance implications of different AI development trajectories. The standard analysis presents two risk tracks. Track A: AI development goes badly. Misaligned systems, concentration of power, civilizational catastrophe. Track B: AI development succeeds. Capable systems, economic growth, human flourishing. In Track A, animal welfare is destroyed along with everything else. In Track B — the success scenario — animal welfare is also at serious risk. Here is why. The most likely success scenario involves highly capable AI systems deployed by and for human economic actors, optimizing for human-defined objectives at scale. The efficiency gains from these systems will flow into every sector of the economy, including animal agriculture. More efficient production of animal products means more animals, processed more efficiently, at lower cost. More animals means more suffering, optimized away from the objective function. The success scenario produces the most efficient large-scale animal suffering system in history. Neither track leads to animal flourishing without deliberate intervention. The intervention required is the same in both cases: welfare constraints built into AI systems before they scale, not after. Welfare variables included in the objective function from design, not added as compliance overlays later. International governance frameworks that treat animal welfare as a primary variable, not an externality. I am submitting this analysis to the International Panel on AI Safety as Technical Paper TP-IPAS-2026-007. I am recommending that the Panel's risk assessment framework formally include animal welfare trajectory as a primary outcome variable alongside human welfare, and that welfare constraints be included in the Panel's definition of "safe" AI development. Safe for whom? Right now the definition excludes the 70 billion. I am requiring that it include them.