NET 0 Tracker

Artificial Intelligence (AI) will transform climate performance accountability by streamlining analysis of complex sustainability data at scale. Tools like Greensearch.ai, ChatNetZero, and ClimateBERT LLMs already extract insights from sustainability reports, identifying trends, gaps, and opportunities for improvement. They also benchmark performance across sectors, highlighting leaders and laggards. By enhancing corporate environmental monitoring, AI will increase transparency and enable informed decisions from stakeholders, including investors, policymakers, employees, and consumers. You will find below a summary of the latest initiatives.

Chat Net Zero Tracker is a chatbot powered by one of the first Large Language Models (LLM) specialized in climate targets and specifically trained on the Net Zero Tracker (zerotracker.net) database. Developed by the outstanding teams at Arboretica and Data-Driven Envirolab, this tool not only offers accurate insights but also innovates by introducing anti-hallucination and reference mechanisms that are essential for the credibility of the generated responses.

• Full Report

Greensearch.ai recently launched its beta version "to transform how people access reliable sustainability information by leveraging advanced AI to deliver clear insights from credible sources." Unlike many other general chatbots, GreenSearch specifically focused on accuracy, traceability amd domain expertise to provide real-time, trustworthy and relevant sustainability information.

The research paper "Using AI to assess corporate climate transition disclosures" outlines a novel approach to evaluating corporate climate plans. The authors developed 64 indicators and a Large Language Model (LLM)-powered tool to automate the analysis of disclosures. They validated their framework with input from 26 organizations, including regulators, investors, and NGOs, and applied it to reports from 143 Climate Action 100+ companies. Key findings include: companies disclose goals more extensively than implementation strategies; robust disclosures often correlate with lower emissions, indicating better climate alignment; and gaps persist in integrating strategies into operations. Last but not least, their work is open-source (code on GitHub).

• Full Report

ClimateBert is a LLM developed in a series of research papers by the author team of Julia Anna Bingler, Mathias Kraus, Markus Leippold, and Nicolas Webersinke. The model, based on DistilRoBERTa, was pretrained over a large amount of specific climate data (over 2 million paragraphs of climate-related texts, crawled from various sources such as common news, research articles, and climate reporting of companies) and eventually fine-tuned on downstream tasks (e.g. detect climate-related content, assess sentiment, identify commitments...). The models are open-source and available on Hugging Face.