Chairman of Agricultural Bank of China, Gu Shu: AI methods can be used to address the risks of AI applications and to combat AI with AI.

On June 18th, at the 2026 Lujiazui Forum's "Sixth Plenary Session: Empowering Financial High-Quality Development with Technological Innovation", Gu Shu, Chairman and Executive Director of the Agricultural Bank of China, stated that in the face of risks in the application of large models, the key is not to eliminate the risks, but to both leverage the benefits of large models and recognize their limitations, while also finding ways to control risks as much as possible. He believes that risks in the application of large models can be prevented and controlled in four aspects: firstly, by implementing tailored strategies to promote scenario adaptation. When applying large models in finance, they are used in different scenarios. For each scenario, a graded control system for the model can be established, matching different scenarios with diverse technical routes and interpretability requirements. Secondly, setting certain standards to constrain and control the large models, as sometimes large models can create illusions or act freely. When applied in different scenarios, parameters and standards need to be established to control their application paths. Additionally, combining them with human input is crucial, as human decision-making plays a key role in the process of large models. Thirdly, using AI methods to address AI application risks, engaging in AI vs. AI to establish a comprehensive defense system, and enhancing the ability of AI to combat risks. Finally, strengthening the internal AI governance system of banks, establishing a comprehensive risk-focused AI governance system throughout the lifecycle to ensure that AI is both used effectively and managed properly, while also retaining innovation space and minimizing the limitations of model black boxes and illusions.