UNESCO Recommendation on the Ethics of AI PA5: paras 84 to 86 Environment and ecosystems: assess and reduce AI's environmental impact, use AI for the planet, and favour efficient methods under precaution

States and companies should measure AI's environmental footprint, direct and indirect, from start to finish, including carbon footprint, energy consumption and raw material extraction, and reduce the impact of AI systems and data infrastructure, with Member States ensuring environmental compliance. Incentives should support rights-based AI for disaster resilience, environmental monitoring and restoration, with participation of local and Indigenous communities and circular economy approaches. AI actors should favour data-, energy- and resource-efficient methods, evidence should show the intended effect or that safeguards justify use, the precautionary principle applies where this cannot be shown, and AI should not be used where environmental impacts are disproportionate.