What happened
Snowflake announced Cortex AI Gateway at Black Hat 2026, a centralized control layer for AI agents that integrates Natoma (an MCP gateway) to govern how agents from multiple sources—including ChatGPT, Claude, and custom apps—access models, data, and enterprise tools. The company also moved several security features to general availability, including Agent Identity and Ransomware Protection via Multi-Party Approval, while placing others in preview.
Why it matters
AI security concerns among tech talent jumped from 17% in 2024 to 48% in 2026, yet 57% of organizations face a capacity gap in security and risk management even as 97% commit to implementing AI. Autonomous agents expand the enterprise attack surface by combining data access, execution, and movement; Snowflake's gateway aims to consolidate fragmented, unmanaged agent sprawl into a single auditable endpoint with fine-grained authorization and cost controls.
What to watch
The gateway offers three core capabilities in varying stages (some in private preview): unified control over agent connections across 100+ MCP servers; real-time audit trails of agent actions and tool calls; and automatic model routing based on cost, latency, and data residency. Snowflake also previewed Data Exfiltration Prevention (DXP), which flags sensitive data fetches triggered by AI agents and unauthorized mass downloads, and CoCo CLI VM Sandbox isolation for client-side AI workflows on macOS.
Summaries like this, in your inbox every morning.
The surge in AI security concerns—from 17% to 48% in just two years—reflects the widening gap between enterprise AI adoption and the security infrastructure needed to protect it. While 97% of organizations commit to AI implementation, 57% report a significant capacity shortfall in security and risk management. Autonomous agents have become a particular vulnerability: by consolidating data access, system execution, and data movement into single profiles, they dramatically expand the enterprise attack surface. Snowflake's announcement addresses this directly by moving security from the periphery—application-layer fixes and legacy monitoring tools—into the core data and control planes.
The Cortex AI Gateway tackles the fragmentation problem: as enterprises adopt agents across multiple models and platforms (ChatGPT, Claude, Azure, Bedrock, and custom applications), decentralized deployment creates unmanaged sprawl, invisible agent activity, and runaway AI costs. By centralizing tool-call governance through Natoma integration, Snowflake enables enterprises to enforce identity, policy, and audit from a single endpoint. This approach extends Snowflake's existing data governance framework to agent traffic, allowing fine-grained authorization across 100+ MCP servers without requiring administrators to manually configure each new agent type.
The move of Agent Identity and Multi-Party Approval to general availability signals that Snowflake views these as production-grade defenses. The addition of Data Exfiltration Prevention—which flags and blocks sensitive data fetches triggered by AI agents—addresses a concrete attack vector. Together, these features reflect a shift in enterprise security thinking: as agents become more autonomous and pervasive, security must transition from reactive monitoring to proactive, context-aware controls embedded directly into data infrastructure.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
