The application of artificial intelligence (AI) is transforming the service delivery of states, policy and environmental system design, and monitoring. The use of AI can enhance the efficiency of administration, refine policy focus and help to achieve Sustainable Development Goals (SDGs) when used wisely. However, discrete and systemic risks include opaque model, biased results, governance asymmetry across jurisdictions and massive ecological footprint of large models and data infrastructures brought about by AI. The given paper creates an interdisciplinary framework, a conceptual one, to introduce transparency, accountability and sustainability into the lifecycle of AI application in the sphere of public governance. The framework combines the AI ethics tools (OECD, EU), sustainability scholarship (SDG frameworks), and critical thinking (decolonial and ecological critiques) and explains the framework by using comparative and contemporary case material. This will consist of two components: (1) three-pillar operational framework of sustainable AI governance, and (2) policy prescriptions (impact assessments of algorithms, AI sustainability certifications, data trusts and participatory oversight) that are both operational and sensitive to global equity.