Bingyang Wu, Yongxi Yi
The governance of AI foundation software ecosystems requires balancing platform scale and independent innovation to ensure technological sovereignty. This paper examines whether tiered open-source mandates-with differentiated requirements for platforms and independents-can optimize ecosystem innovation. We develop and solve a tripartite Stackelberg differential game model incorporating differences in compliance costs, ecosystem dependence, and appropriation capabilities. Analysis shows that tiered mandates outperform uniform ones when heterogeneity exceeds threshold levels, optimally imposing stricter rules on platforms (θD∈[0.60,0.75]) and offering flexibility to independents (θI∈[0.25,0.40]). Ecosystem health improvements range from 15-40%, peaking under high compliance cost asymmetry (λD/λI>2.0). Key boundary conditions are identified, and an adaptive decision framework is proposed. This research formalizes differentiated mandates as an endogenous policy variable, quantifies dominance conditions, and provides actionable thresholds, demonstrating that calibrated differentiation based on structural heterogeneity can optimize innovation in AI infrastructure governance. Throughout this paper, "health improvement" refers specifically to the ecosystem architect's composite health function H(KD,KI), a governance metric of scale, diversity, and balance, not a full measure of social welfare.