Saviynt and Snowflake add runtime access controls for third-party AI agents

Saviynt has announced an integration with Snowflake designed to reduce identity risks when third-party AI agents access enterprise systems. By bringing Saviynt’s Access Gateway into Snowflake’s security environment, the solution enables organisations to connect, govern and audit third-party agents across business systems and platforms. Under this setup, access decisions are evaluated at runtime based on an agent’s intent, context, permissions and risk.

AI agents can make autonomous decisions, access sensitive data and invoke external tools. Saviynt shares that these interactions create security and governance requirements that conventional identity controls designed around human users may struggle to address.

Access decisions are evaluated at runtime

Saviynt’s Access Gateway is designed to assess what an agent is attempting to do when it requests access to enterprise resources. The system evaluates the agent’s intent and context alongside its permissions and associated risk before determining whether an action should be authorised. This structure aims to restrict an agent to the actions required for a specific task, reducing the risk of excessive privileges being used across connected systems.

Akhil Kapoor, Senior Vice President, Business Development at Saviynt, described the distinction between human and agent access controls. He stated, “Traditional identity security models were built for human users, not AI agents, creating a critical gap in enterprise security. Our integration with Snowflake addresses this challenge through Saviynt’s Access Gateway, which evaluates agent intent, context, permissions, and risk at runtime to ensure AI agents are only authorised to perform the actions required for a specific task. It prevents privilege misuse and anomalous activity before it can impact the business.”

The integration also provides governance and auditing capabilities for third-party agents operating across platforms. This gives organisations a practical way to apply access controls as agents interact with multiple business systems instead of treating each connection independently.

Snowflake extends controls to third-party agents

Snowflake is incorporating Saviynt’s technology into its security ecosystem as part of its approach to governing third-party agents. The companies are focusing on what they describe as trusted agent interoperability, where autonomous agents can work across enterprise systems while remaining subject to defined access permissions and governance controls.

Mayank Upadhyay, Chief Security and Trust Officer at Snowflake, said the integration is intended to provide a foundation for running autonomous workflows across connected systems. He noted, “Securing the agentic enterprise requires an entirely new model of trust, visibility, and control. Through our integration with Saviynt, we are establishing a secure foundation for agent interoperability, equipping organisations with the precise access capabilities and trust required to run autonomous workflows safely.”

The integration centres on controlling what third-party agents can access across enterprise environments. Evaluating agent actions at runtime allows organisations to maintain consistent governance across all the systems and platforms in which these agents operate.