A groundbreaking, one-of-a-kind artificial intelligence model developed to map and track carbon emissions across global production chains, consumption patterns, and natural carbon sinks has been unveiled by a team of Chinese researchers, a development that experts say could reshape dynamics in international climate negotiations and rewrite how global emissions accountability is calculated.
The innovative large language model was publicly launched on Wednesday by the Shanghai Advanced Research Institute (SARI) under the Chinese Academy of Sciences, rolling out at a pivotal moment when China works to support domestic enterprises in hitting ambitious carbon reduction targets while cementing its position as a technical leader in global climate governance frameworks.
Built as a large language model trained on petabytes of structured and unstructured environmental data, the system boasts 32 billion parameters — the core AI building blocks that act like neural synapses, enabling the tool to detect complex emission patterns and generate accurate, data-backed predictions. To handle the multifaceted nature of global carbon tracking, the model integrates five specialized artificial intelligence sub-programs, called intelligent agents, each tailored to a distinct critical task.
These specialized agents cover a wide range of use cases: digital simulation to identify the most energy-efficient operational configurations for industrial factories, cross-border carbon transfer tracking that maps how embodied carbon moves between countries through global trade, full life cycle assessment that calculates a product’s total environmental footprint from raw material extraction through end-of-life disposal. The system also includes a natural carbon source accounting agent to quantify carbon sequestration from ecosystems like forests, and an uncertainty analysis agent to validate the reliability and consistency of all output data.
Gao Yunhu, a lead researcher at SARI, described the AI as a specialized “carbon accounting butler” that outperforms legacy carbon tracking methods by a wide margin. Traditional carbon accounting workflows are notoriously slow, labor-intensive, and costly for businesses, but the new model enables real-time simulation of production processes to help companies identify the most cost-effective pathways to cutting emissions.
Zhang Xian, director of the Division of Global Environment at the Administrative Center for China’s Agenda 21, echoed this assessment, noting that conventional accounting methods not only drain time and resources but also make it nearly impossible for enterprises to measure emissions accurately across every stage of a product’s supply chain. Unlike these outdated approaches, the new AI tool can conduct a full life cycle assessment starting from the extraction of raw materials, turning what was once a costly regulatory burden into a competitive advantage for businesses by enabling targeted deployment of emission-cutting technologies.
Lai Xiaoming, chairman of the Shanghai Environment and Energy Exchange, explained that by standardizing emissions quantification across entire industrial and supply chains, the model improves market monitoring, emissions quota verification, and climate policy impact assessment. It also provides robust, reliable technical infrastructure to support global green trade and transparent carbon pricing systems, he added.
The launch comes at a particularly critical juncture for Chinese exporters, who now face new carbon-based import taxes under the European Union’s Carbon Border Adjustment Mechanism (CBAM), which imposes a price on carbon embedded in carbon-intensive goods including steel and cement entering the EU bloc.
Mi Zhifu, a professor of climate change economics at University College London, pointed out that the EU currently relies on standardized “default values” to estimate emissions when an importing company cannot provide independently verified emissions data. For many Chinese products, most notably steel, these default values are often significantly higher than the actual emissions generated during production, incorrectly painting Chinese goods as more carbon-intensive than they truly are and exposing exporters to unnecessary extra taxes. By generating independently verifiable, granular emissions data, the new AI model helps Chinese firms avoid these inflated tax assessments, especially in key CBAM-regulated sectors including steel, cement, hydrogen, electricity, and fertilizers.
One of the model’s most distinctive contributions to global climate accounting is its core focus on consumption-based emissions accounting, a departure from the territorial-based standards that dominate current international frameworks. Under current common standards, emissions are attributed entirely to the country where production takes place. The SARI model, by contrast, tracks “embedded carbon” — the total carbon footprint hidden within finished products traded across borders — to recognize that consuming countries share equal responsibility for the emissions generated during production.
As an example, Wei Wei, vice-president at SARI, cited China’s exports of renewable energy technology. In 2024, the manufacturing of Chinese-made wind turbines and solar panels generated approximately 2 million metric tons of carbon emissions. Over the operational lifespan of these products, however, they will help countries around the world cut more than 350 million metric tons of carbon emissions, a net climate benefit that is not reflected in traditional territorial accounting.
Traditional accounting frameworks attribute all emissions from manufacturing for export to China, which obscures the emissions responsibility of developed countries that consume these traded goods, Zhang noted. By quantifying these unrecognized “carbon leaks” embedded in global trade, Chinese climate officials and researchers believe the model provides a more scientifically sound foundation for future international climate negotiations, enabling more equitable allocation of global emissions reduction responsibilities.