New tool helps Etsy peer into the ‘black box’ of AI emissions

KUALA LUMPUR, 30 SEPTEMBER 2026 —

Sustainable AI Group has launched CLE — Closed‑model Latent Energy Estimation Range — a free, open‑source tool that translates the number of tokens processed by enterprise artificial‑intelligence models into estimates of energy use and greenhouse‑gas emissions, the company said on Thursday.

The tool is intended to give organisations a way to quantify the carbon footprint of AI workloads that are otherwise hidden inside proprietary models, and it will become more precise as additional firms adopt the methodology, Sustainable AI Group added.

CLE works by converting a company’s token consumption – the text that large‑language models from providers such as Anthropic, Google, Microsoft and OpenAI handle – into energy and emissions figures. The conversion relies on a dataset that draws on publicly available information for more than 20 open‑source models, creating proxy values for the closed‑source models that many businesses licence.

As model providers disclose actual measurement data, those figures can replace the proxy estimates without requiring users to rebuild their accounting approach, Sustainable AI Group said.

The methodology was co‑developed with e‑commerce platform Etsy, which is among the first customers to test the tool, together with private‑equity firm Hg and tax‑software company Sovos. “When new environmental challenges emerge, we don’t wait for the industry to give us new answers,” said Chelsea Mozen, head of impact and sustainability at Etsy. “We test new approaches, learn what works and share those insights so others can move forward with us.”

Etsy has previously built its own system for measuring emissions from cloud‑computing services and published the approach for other companies to use. It is one of the few firms that separately reports emissions from digital operations as part of its broader disclosures for purchased goods and services.

While CLE is offered at no cost, companies may also licence specific datasets for integration into their existing carbon‑accounting platforms on an ongoing basis – a route Etsy plans to pursue.

The tool also enables comparative analysis of AI model performance. Sustainable AI Group’s technical analysis found that model selection for particular tasks can alter energy consumption by more than 30 percent, and that lower‑cost models often exhibit higher estimated energy needs. “In the case of a sustainability professional, pretty much every one they use would apply,” said Boris Gamazaychikov, co‑founder and CEO of Sustainable AI Group.

CLE builds on the co‑founders’ earlier work on the AI Energy Score, a resource that calculated the energy and emissions of widely available AI models. The new framework extends that effort to the closed models that dominate corporate AI deployments, using the open‑model data as a basis for estimation.

Sustainable AI Group is also preparing an integration of CLE with Watershed, a sustainability‑platform used by roughly 800 companies. Watershed is developing a complementary framework for accounting AI emissions, and the two organisations will collaborate on embedding CLE data into the platform. “We want this data to go into the places where sustainability professionals are already doing their accounting,” Gamazaychikov said.

The launch arrives as corporations across Malaysia and the wider region grapple with the rising energy demands of AI adoption. By providing a transparent, comparable metric for AI‑related emissions, CLE offers a practical step for firms seeking to incorporate digital‑operation footprints into their sustainability reporting and to meet emerging regulatory expectations on carbon accounting.