Snowflake CoCo expands AI governance with new enterprise controls

Snowflake is building on its July announcement about CoCo's ability to scale enterprise AI with trust. The new controls address a common friction point for platform teams: giving developers broad access while keeping risk measurable. Before these changes, teams often limited CoCo to narrow groups, sandbox schemas, or manual approval processes because the productivity gains were clear but the risks were harder to quantify.

The expansion moves governance from a focus on cost to a more complete framework covering access, behavior, and tooling. Per-user quotas handle spend, while managed device management pushes organization-wide policy across every CoCo installation. Agent profiles apply defaults by role, and restricted session scope limits what SQL an agent can run until an appropriate role is active. Each control answers a specific security review question, which could make it easier for administrators to open CoCo broadly.

For builders, the intended benefit is that governance becomes invisible. Agent profiles load the right model and skills by role, and governed external connections appear automatically without manual configuration. The company frames these controls as a way to give more developers production access, not less. That said, the article is a forward-looking product announcement, so actual availability and feature details may differ from what is described here.