QuestionQ73

VerifiedSecuring and governing agentic workflows

You are setting up security and monitoring controls on Agent Gateway for an enterprise financial agent. The agent calls external third-party partner APIs through Agent Gateway.

You need to configure Agent Gateway to enforce these policies:

  1. Outbound traffic passing through Agent Gateway must be continuously inspected to detect and block unintentional exfiltration of sensitive data in tool payloads.
  2. Egress traffic metrics, including per-agent token usage, request latency, and HTTP response codes, must be aggregated in real time for centralized security monitoring and cost allocation.

What should you do?

Choose two
Explanation

Outbound content inspection can detect and block or redact sensitive data before supported payloads reach their destinations. Agent Gateway supports this through Model Armor configured with Sensitive Data Protection settings. Centralized telemetry supports monitoring and cost analysis; gateway logs can be routed from Cloud Logging to BigQuery. (docs.cloud.google.com)

The choices are approximate: the documentation does not establish a direct Sensitive Data Protection filter on gateway routes or route-level exporters that supply all the listed per-agent token metrics. Model Armor also sanitizes only supported egress payloads, so coverage of arbitrary partner APIs must be verified. (docs.cloud.google.com)

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!