The Bank for International Settlements (BIS) has completed the first stage of a new project that is using generative AI (genAI) to carry out climate risk analysis.

Project Gaia aims to help analysts search corporate climate-related disclosures and extract data quickly by using large language models (LLMs).

The first phase surveyed climate risk experts from central banks and supervisory authorities; designed a solution that addresses the requirements articulated by these experts; and delivered a proof of concept (PoC) demonstrating the technical feasibility of the concept.

BIS said that the PoC enabled the comprehensive analysis of climate-related indicators through automating information extraction and offered "harmonised metrics", despite the heterogeneity of naming conventions and definitions across different jurisdictions.

"The combination of semantic search together with iterative and systematic LLM prompting enabled Gaia to overcome differences in disclosure frameworks," it continued. "This offers much-needed transparency and comparability of climate-related information."

However, several technical issues were unearthed when LLMs were integrated into an application for data extraction. These included long response times, randomness in responses, and hallucinations.

But BIS outlined a set of design choices that allowed the PoC to overcome these challenges.

"Thanks to its flexible design, the platform is relevant in a broader context than climate-related data analysis, showcasing the value of AI-enabled applications for central banks and the financial sector," said BIS. "Generative AI promises to change the way we work in the future, and Project Gaia is one of the first comprehensive studies investigating how this can be done in practice.

---