Introduction

As climate change intensifies, financial losses linked to climate risks are becoming harder to insure or hedge - partly because climate risk is still poorly defined and measured. This lack of clarity is concerning, as it can lead to the buildup of financial risk in specific areas, potentially threatening the stability of the broader financial system.

We will investigate the possibility of using Artificial Intelligence (AI) to analyse company reports, climate reports, and more high-frequency data (e.g., daily news) to obtain representations of firm-level climate risk exposure. Recent developments in Natural Language Processing (NLP), Computer Vision and Deep Learning models, such as transformers, have made it possible to create much better text (or image) representations that could be deployed in a wide variety of tasks. By leveraging these new techniques, we aim to develop climate risk metrics, ultimately improving how markets assess and price climate risk.

In addition, we integrate these AI-based climate risk metrics into traditional asset pricing frameworks to test their financial relevance. Finally, we extend the analysis to a macro-financial level, exploring how climate risk propagates through the financial system.