Snowflake Chief Security and Trust Officer Mayank Upadhyay said security must be built into the data platform and become part of normal workflows. He said Snowflake is approaching AI security in three layers: protecting large language models, managing agent identity and permissions, and securing the data layer. He highlighted the Horizon Guardrails product and the Natoma acquisition, and stressed fast feedback loops, data sovereignty and zero data retention.

Pulmuone’s supply chain management executive Ahn Soon-sik (안순식) said the food company is applying AI on Snowflake’s platform to build an SCM intelligence project and that creating AI-ready data is critical. He said short shelf life makes demand errors costly and overseas operations add complexity. Pulmuone spent over six months cleansing and standardising data and built a two-layer knowledge base. Internal tests showed AI analysis matched Snowflake data, and a modelling shift was completed in three weeks using Coco.