Establishing global governance of artificial intelligence (AI) is becoming an increasingly pressing challenge to ensure the provision of global public goods and to mitigate harmful effects on societies and the planet. Current debates around AI take various forms, follow diverse narratives, and centre variously on economic, social, environmental, or safety aspects. Here, we make three contributions. First, we classify risks and challenges of AI across the social, planetary, and safety domains. Second, we show that AI should be governed as a global commons, requiring coordinated interventions across all three domains, reflecting relevant inter-domain feedback loops, and root drivers, such as the pursuit of monopolistic AI power and the AI-infused media environment. Third, we identify data, energy, and compute as relevant regulatory dimensions across social, planetary, and safety domains. We conclude by emphasising the importance of limiting agentic AI, incentivising depolarising algorithms on social media, and setting AI dynamics within the context of global equity.© 2026 The Author(s)
The increasing deployment of large language models (LLMs) in natural language processing (NLP) tasks raises concerns about energy efficiency and sustainability. While prior research has largely focused on energy consumption during model training, the inference phase has received comparatively less attention. This study systematically evaluates the trade-offs between model accuracy and energy consumption in text classification inference across various model architectures and hardware configurations. Our empirical analysis shows that in some contexts the best-performing model in terms of accuracy can also be energy-efficient. While LLMs tend to consume significantly more energy than traditional machine learning models, they show the same or even lower levels of accuracy in our zero-shot classification setting. We observe substantial variability in inference energy consumption (mwh to kWh), influenced by model type, model size, and hardware specifications. Additionally, we find a strong correlation between inference energy consumption and model runtime, indicating that execution time can serve as a practical proxy for energy usage in settings where direct measurement is not feasible. Our findings demonstrate that energy efficiency and accuracy represent distinct evaluation dimensions that do not necessarily align. We argue that sustainable AI development requires systematic evaluation of both performance and resource efficiency. © The Author(s) 2026