Snowflake Launches Cortex AI Gateway for Agent Control and Cost Management
Snowflake has introduced Cortex AI Gateway, a new centralized control layer designed to manage how AI agents access enterprise data, tools, and models. Announced on Tuesday, this gateway aims to prevent unauthorized access and control escalating costs associated with AI deployments. The company also revealed initial security integrations with identity vendors 1Password, Aembit, Linx Security, SailPoint, and Saviynt, forming an unusual alliance focused on establishing a shared trust model for autonomous agents.
This move signifies Snowflake's ambition to become more than just a data repository, positioning itself as a crucial control plane for AI operations. Mayank Upadhyay, Snowflake's chief security and trust officer, emphasized that the future of AI relies on secure interoperability rather than closed ecosystems, which he believes lead to fragmentation and limit innovation. He argued that traditional security models, built for human actors, are inadequate for AI agents operating at machine speed, exposing pre-existing blind spots in enterprise visibility and amplifying risks.
Cortex AI Gateway will govern both Snowflake's internal agents and third-party agents, centralizing access policies, authentication, permissions, and audit logging. It also addresses the growing concern of runaway AI spending by providing IT and finance teams with a unified view of AI consumption, attributing costs to specific agents or workloads, and enforcing spending limits. The gateway builds upon Snowflake's acquisition of Natoma, a startup specializing in centralized gateway technology for tool-call level security and auditing. The system employs dual attribution, logging both the agent's identity and the human who authorized the task, ensuring task-scoped access and complete auditability.
Snowflake's introduction of Cortex AI Gateway addresses a critical inflection point in enterprise AI adoption: the need for robust governance and cost control as autonomous agents become more prevalent. The announcement highlights a fundamental shift in security paradigms, moving from human-centric access controls to agent-centric identity and authorization. By centralizing policy enforcement and cost attribution, Snowflake aims to mitigate risks associated with AI's speed and complexity, such as unintended access amplification and spiraling operational expenses. The collaborative approach with multiple identity vendors suggests a recognition that no single entity can solve the multifaceted challenges of AI agent security and management, fostering an ecosystem-based trust model. This strategic positioning anticipates the growing demand for platforms that can manage the dual imperatives of AI innovation and enterprise-grade security and financial oversight in the coming decade.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.