Exclusive: Snowflake invests in Bedrock Data to strengthen agent AI system governance

Data governance, security and management startup Bedrock Labs Inc. is deepening its integration with Snowflake Inc.’s AI Data Cloud after securing a strategic investment from the cloud data warehousing giant.

The new partnership announced today integrates Bedrock’s artificial intelligence-driven data classification and governance capabilities with Snowflake Horizon, a service within the AI ​​Data Cloud that helps organizations manage their data, applications, and AI models.

Additionally, Bedrock announced a new integration between ArgusAI and Snowflake’s Cortex AI. It uses AI agents to perform tasks such as text processing, summarization, and predictive analytics on behalf of unskilled workers. This allows companies to do this while strengthening governance and reducing risk.

Bedrock said the integration is timely as organizations are under pressure to implement AI into their data workflows. The problem is that 79% of companies struggle to classify sensitive data used in AI systems. To do this, you need exactly the petabyte-scale data discovery, classification, and qualification analysis that Bedrock provides.

Mapping data confidentiality

Co-founder and CEO Bruno Kurtic said the company is particularly enthusiastic about integrating with Snowflake as it is one of the most popular data platforms for modern enterprises. “Securing data across both traditional analytical workloads and emerging AI applications is a fundamental requirement for an enterprise’s AI strategy,” he said. “Snowflake’s investment confirms that data-centric governance is not just a nice-to-have, but a prerequisite for confidently deploying AI.”

Starting today, Bedrock is integrating its proprietary Metadata Lake platform with Snowflake Horizon. Metadata Lake is a continuously updated graph knowledge base that maps all aspects of enterprise data according to sensitivity, lineage, entitlement, access patterns, and business context. With this integration, we can now map all of this directly to the Horizon platform. It serves as a single source of truth for data sensitivity and risk context, and enables the use of AI agents across the Snowflake platform.

Customers benefit from continuous visibility into their Snowflake environments. Metadata Lake allows you to automatically discover and classify all sensitive information stored within it, including personally identifiable information and financial data. For each dataset, Bedrock assigns impact scores to related schemas and tables based on how sensitive the files are. Customers can then implement their own appropriate security controls.

Additionally, Bedrock can accurately identify who has access to the types of data stored within Snowflake by mapping entitlements across users, accounts, roles, and AI agents. This is done by using Snowflake’s native tagging capabilities to label information at the database, table, and column level based on sensitivity. You can then implement access controls within your Snowflake environment, while the Horizon Catalog keeps everything up to date in real-time.

AI agent data access control

Bedrock’s second integration push combines ArgusAI with Snowflake’s Cortex AI to enable inventory and cataloging of Cortex agents, allowing enterprises to map the data they are granted access to through Cortex Search and Cortex Analyst. Bedrock plans to introduce this integration at the RSA Conference in San Francisco next week, and it will be available shortly thereafter.

This will be helpful because traditional data security posture management tools were built before the era of AI agents, Krutic said. Although it can discover sensitive data, it lacks the ability to map agent relationships, access paths, and permissions.

ArgusAI does all of this, enabling businesses to create unified exposure maps that help them understand and contain the risks associated with their agent systems. ArgusAI accomplishes this by leveraging Bedrock’s data bill of materials, a continuously updated inventory of data assets linked to AI systems. Organizations can then discover the Snowflake Cortex Search service, identify the datasets indexed by the service, and gain visibility into which datasets are accessible to AI search applications. You can also associate Cortex with role-based access controls to specify which AI agents and applications can access sensitive data indirectly through AI search tools. Teams can then flag any entitlement gaps they discover, remediate access paths, and ensure the appropriate controls are in place.

“When companies deploy AI, the risk is determined by what those systems have access to,” Krutic said. “You can’t manage agents unless you know what data they can access, through which MCP servers, and with what credentials.”

Harsha Kapre, head of Snowflake Ventures, said strong governance minimizes risk and is key to AI adoption. “Bedrock Data’s integration with Snowflake Horizon and Snowflake Cortex AI will help our joint customers accelerate their AI efforts while maintaining security and compliance,” he said.

Image: SiliconANGLE/Gemini