209-7 GeoGPT: Building a Governance Framework for Trustworthy AI in Geological Sciences

Session: Deep-Time Earth and the AI Revolution

Presenting Author:

Author:

Abstract:

Artificial intelligence (AI) is transforming scientific discovery, with large language models (LLMs) providing unprecedented capabilities for synthesizing and interpreting complex geoscientific data. However, integrating AI into the Earth sciences raises critical concerns about data integrity, transparency, ethical oversight, and public trust. Addressing these challenges is essential to ensure responsible and safe deployment in research contexts.

This presentation introduces GeoGPT, an AI platform developed by Zhejiang Lab to support applications including geological mapping, fossil identification, predictive modeling, and energy forecasting. Recognizing the societal and scientific implications of such technologies, GeoGPT has established an independent Governance Committee to guide responsible AI development and deployment. Guided by principles of openness, transparency, scientific integrity, and ethical accountability, the committee provides strategic direction, ensures regulatory alignment, and builds trust across scientific and public domains.

GeoGPT's governance framework draws on international models such as UNESCO's AI Ethics Guidelines, the OECD AI Principles, and evolving regulatory policies in China, the United States, and the European Union. Key initiatives include risk and bias assessment protocols, mechanisms for public and expert engagement, and a roadmap for compliance with emerging AI standards in science. These measures aim to maximize scientific and societal benefits while mitigating unintended consequences.

By sharing GeoGPT's experience, this work offers practical insights into governance structures that foster innovation while upholding responsibility, providing a potential model for the broader geoscience community in the era of intelligent systems.

Geological Society of America Abstracts with Program. Vol. 57, No. 6, 2025

doi: 10.1130/abs/2025AM-11128

© Copyright 2025 The Geological Society of America (GSA), all rights reserved.