Sumsub Launches KYA Module to Control AI Agent Access

The KYA Module for MCP addresses a specific concern in AI adoption: the potential for autonomous agents to access sensitive company data without sufficient oversight. By introducing a filter that removes personally identifiable information and a configurable rule engine, Sumsub aims to provide businesses with more granular control over what their AI agents can see and do.

The system's ability to trigger real-time liveness checks for critical actions, such as approving high-value invoices, introduces a human-in-the-loop mechanism. This approach attempts to balance the efficiency benefits of AI agents with the need for accountability and security, particularly as agents become more integrated with financial and communication systems. An auditable record of agent activities also supports compliance and incident review.

The module's foundation in the Model Context Protocol (MCP), an open standard, suggests an effort to ensure interoperability across different AI models and external data sources. While initially supporting Amazon Bedrock AgentCore Gateway, planned support for additional MCP gateways could broaden its applicability, potentially making it a more versatile tool for companies using diverse AI agent platforms.