Since the 2015 Paris agreement, there has been a greater focus on sustainability in finance and economics, both in research and market practice. This has been accompanied by international financial regulators and supervisors working to develop a new financial architecture that properly incorporates and manages climate-related opportunities and risks.

There are important challenges in scaling up green finance, such as the still limited reliability of climate data and the statistical complexity of modelling the non-linear behaviour of climate change and its impact on the economy, which is twofold due to the existence of a feedback loop. Increasingly, these issues justify the use of new technologies such as artificial intelligence and machine learning to overcome operational barriers that could prevent new green financial products and markets from existing.

In 2021, the Bank for International Settlements created the Innovation Hubs to experiment within the international community of central banks and track technological trends and developments in green finance. The goal is to develop public goods with different technologies, such as machine learning, natural language processing or blockchain. Currently, the hubs have several projects underway that are using these technologies to overcome climate-related challenges.

Project Gaia is exploring the use of large language models to automate the collection and management of climate-related information. Project Genesis 1.0 and 2.0 work on the development of digital green bonds, while Project Symbiosis investigates how to improve supply chain sustainability disclosures using AI and big data. Project Viridis geo-locates the exposure of banks to extreme weather events.

To further investigate how and when researchers are using machine learning models in green finance, we ran a systematic review of studies using this technology in a wide variety of topics related to climate change and economics. Our results support the relevance of this technology as a key lever in green finance. We reviewed a wide variety of methods, finding that the most complex ones, like artificial neural networks, do not lead in all thematic areas, as either the datasets available do not have the proper characteristics, or policy guidelines require more interpretable model specifications.

Leveraging technology

On this front, international financial supervisors and regulators do not only experiment or research inside their institutions or across themselves, but also promote innovation within the private sector. Currently, the G20 Techsprint 2024 – Technology for the Planet – is the fifth edition of a global hackathon competition focused on solving sustainable finance problems.

The main areas where technology might underpin further developments in green finance are the potential use of satellite data for nature finance and physical risks evaluation, the use of LLMs for transition plans analysis (e.g. Chatreport) and improving scope 3 carbon emissions measurement through machine learning or external assurance of disclosures. Broadening the scope, some major jurisdictions are running similar initiatives, like the Green Fintech Challenge in the UK, or the Monetary Authority of Singapore’s Global Hackcelerator, which has also turned its attention to green finance.

However, it must be acknowledged that AI is an energy-consuming technology, which is a concern that has given rise to the notion of ‘green AI’, questioning its environmental impact on the planet. A successful use of this technology should be deemed to comply with the principles of sustainability, but also with the guidelines of responsibility and ethics in AI development to strike a balance between the risk and opportunities of this innovation for a sustainable planet.