Sumsub adds request and data-access controls to Know Your Agent

Sumsub announced on 19 October a Know Your Agent module for controlling AI-agent requests and access to information. Initial support is for Amazon Bedrock AgentCore Gateway, with further gateway connections planned.[1]

The company describes real-time policy checks for connected tools and data, including sensitivity rules and handling of personally identifiable information. Higher-risk requests can trigger further verification. This adds action-level decisions to the product’s earlier identity and configuration functions.[1]

Identity supplies context; policy determines access

The January release tied an agent’s activity to a verified person. October’s module considers the requested action and data as well as that identity. Sumsub says it can retain activity records and request measures such as liveness checks or additional information when warranted.[2][1]

For example, an agent retrieving a broker’s public product information and the same agent requesting personal client records require different permissions. A useful decision record would identify the action, applicable policy, outcome and any escalation. This is an illustrative control design.

Separate configuration from approval

Sumsub’s June product lets agents use a business’s policies to build verification configurations, including questionnaires, risk settings and workflows, within its dashboard. That creates a related change-control task: organisations must decide who may alter a policy and who may approve access under it.[3]

The commercial proposition is continuous control around connected automation. Its operation depends on the configured rules, integrations and exception handling. The announcement supplies a vendor description of capabilities; measured error rates and deployment outcomes were not published.[1]

Report an error or submit an update