Algorithmic Sustainability: Ethics, Governance, and Global Policies for AI-Driven Environmental Decision-Making

In recent years, artificial intelligence has become embedded in environmental decision-making, shaping how climate risk is zoned, how land use is planned and managed, and how regulatory oversight and energy-related decisions are carried out. Despite this expansion, discussions surrounding the use of AI in decisions related to sustainability often focus on performance measures, with limited attention given to broader institutional and environmental implications. Such accounts frequently sidestep questions of governance legitimacy while underestimating the environmental burdens associated with computational processes and the infrastructure that supports them. This paper develops algorithmic sustainability as a governance framework oriented toward public policy in contexts where artificial intelligence informs environmental decision-making. The concept is defined through the simultaneous alignment of three conditions. These include ecological effectiveness assessed across the full lifecycle of AI systems, institutional accountability anchored in oversight that can be enforced in practice, and ethical legitimacy grounded in freedom, justice, and the possibility to contest decisions. Rather than treating these dimensions as separable, the framework assumes that sustainability claims weaken when any one condition is absent. The research methodology adopts a framework-development approach supported by a qualitative comparative review. The review integrates scholarship on climate impact pathways with ethical and political analyses of algorithmic authority, while also drawing on governance instruments found in global normative frameworks, regional regulatory models, and organizational practice. Through this synthesis, the paper produces two outcomes. One is a four-domain ethical risk register that consolidates epistemic and technical concerns, risks tied to justice and political economy, issues of accountability and legitimacy, and impacts associated with the environmental footprint of AI systems over time. The second outcome is a governance toolkit that translates algorithmic sustainability into practice through proportional risk tiering based on decision criticality, requirements for documentation and auditability, a tiered Environmental AI Impact Assessment, standardized disclosure of environmental footprints, procurement-based leverage, and enforceable mechanisms that allow contestation and remedy. The analysis shows that environmental AI governance remains institutionally fragile when sustainability evaluation is disconnected from transparency obligations, challenge pathways, and distributive accountability as they operate in practice.