Snowflake CoCo expands AI governance with cost, access, and tool controls

In July, Snowflake outlined a vision for governing AI costs, grounding AI in enterprise context, and bringing trusted AI to work. This launch delivers on that vision by expanding governance beyond cost to access, behavior, and tooling. Administrators can now set per-user quotas, enforce organization-wide policies via managed settings, define defaults with agent profiles, and constrain agent actions with restricted session scope. These controls address common security review questions, potentially reducing the need for sandboxes and manual approvals. For builders, the governance layer is designed to be invisible in normal use, surfacing only when an action would be problematic. External tool access through the Cortex AI Gateway adds another layer of control and auditability, built on Snowflake's Natoma acquisition. This holistic approach aims to balance security with developer productivity, enabling broader AI adoption within enterprises.