Contesting AI infrastructure for just sustainabilities: reframing large language models in environmental justice and local policy

Large Language Models (LLMs) are increasingly proposed as tools to support sustainability policy, environmental planning, and local governance. Yet their development and deployment often reflect globalised knowledge hierarchies and corporate control, posing risks of epistemic exclusion, platform capture, and digital inequity – particularly in marginalised communities. This Comment critically engages with the political and environmental implications of LLMs, arguing that they must be reframed not as neutral innovations but as contested infrastructures shaped by local knowledge, struggles, and justice-centered values. We explore how LLMs can be localised, governed democratically, and co-produced with communities to support just sustainabilities. Through a framework grounded in epistemic equity, pluralistic design, and participatory AI governance, the article proposes pathways for transforming LLMs into tools of collective empowerment rather than techno-solutionist displacement. Reimagining LLMs as algorithmic commons opens possibilities for inclusive, place-based, and emancipatory environmental futures.