Snowflake Bolsters AI Security with Natoma Acquisition
Cloud data platform leader Snowflake has acquired Natoma, a startup specializing in Model Context Protocol (MCP) technology. This move strengthens enterprise control over increasingly complex AI agent deployments.
As organizations transition from experimental AI workflows to real-world applications, the need for robust governance frameworks is growing. MCP enables secure connections between AI agents and various business systems—including SaaS applications, cloud environments, and on-premises infrastructure.
Natoma’s platform provides tools for accessing data securely, managing user identities, implementing policies, and auditing AI activity within Snowflake’s ecosystem. This integration allows organizations to govern access across platforms like Cortex Agents, Snowflake Intelligence, and Cortex Code—ensuring compliance and reducing security risks.
The Growing Importance of AI Governance
Experts note that while MCP provides a technical foundation for connecting AI agents, comprehensive governance is essential. HFS Research CEO Phil Fersht warns that without proper controls, MCP deployments could create “shadow AI” environments where unauthorized access and activity occur outside IT oversight.
“CIOs must manage not only who can view data but also what information agents can access, which systems they can interact with, and what actions they’re authorized to perform,” Fersht emphasizes.
Robert Kramer of KramerERP adds that MCP alone isn’t sufficient: “It’s a protocol, not a governance model. While it standardizes connections, risks remain if access is overly permissive or controls are inadequate.”
Enterprise Readiness for AI Agents
While the acquisition positions Snowflake to meet growing demand for AI governance solutions, industry analysts suggest many organizations aren’t fully prepared for this new era of technology.
“Companies want the benefits of AI agents—increased productivity and contextual awareness—but their governance frameworks haven’t kept pace,” notes Fersht. He cautions that without proper controls, AI agents could expose sensitive data or bypass existing security measures.
Key areas CIOs should focus on include:
- Identity-based access management
- Least privilege principles (granting only necessary permissions)
- Comprehensive audit trails for all AI activity
- Human review processes for high-risk actions
- Data loss prevention controls
- Clear accountability frameworks when AI makes errors