Snowflake expands CoCo AI governance with access and behavior controls

Snowflake's announcement builds on a July blog post about scaling enterprise AI with trust, structured around three themes: managing costs, grounding AI in enterprise context, and embedding AI into existing workflows. Today's expansion addresses a core tension in enterprise AI adoption: teams want broad developer access to AI agents for productivity gains, but security and platform teams need visibility and control to manage risk. The new controls answer this by distributing governance across four dimensions—cost, data reach, behavior, and external integrations—each enforced at session initiation rather than monitored after the fact.

Per-user quotas are the first and simplest control, now live across all CoCo surfaces. The three additional layers tackle governance depth: organization-wide policy (MDM) ensures consistent rules whether developers use the web interface, CLI, or desktop; agent profiles handle role-based defaults without per-developer setup; restricted session scope (RSS) directly answers the "blast radius" question by constraining an agent to SQL permissions tied to active roles, not implicit user credentials. The introduction of Tools by Cortex AI Gateway—built on technology from the Natoma acquisition—signals a strategic pivot toward centralized, auditable external tool governance, addressing a gap where teams previously had few controls over MCP server access.