AI Governance / Platforms / Security
Google adds AI agent observability across BigQuery and Looker, logging prompts to customer-owned tables
Telemetry from Google's own data agents and custom ADK or LangGraph agents lands in BigQuery, with opt-in conversation logging that skips turns touching masked or row-secured data.

Google has announced expanded observability and evaluation features for AI agents across BigQuery and Looker. They cover its own Conversational Analytics agents and custom agents built with ADK or LangGraph. Telemetry is written to BigQuery tables the customer owns. The Conversational Analytics API, now generally available, carries admin audit logs, CMEK, VPC Service Controls and data residency options.
Google says operational metrics and trace spans leave out prompts, responses and generated SQL. Agent owners can opt in, agent by agent, to conversation logging (generally available) into an agent_events table, which inherits the dataset’s IAM permissions, CMEK and VPC Service Controls. A fail-closed check before each turn skips logging whenever a knowledge source uses row-level security, column policy tags or dynamic data masking; if logging itself fails, the chat carries on.
Cost controls include per-query byte-scan caps. Daily token quotas per project, user or agent are in preview. So are native BigQuery agent evaluations, alongside Google’s open-source Prism benchmarking tool.
Editor’s note: This article relies solely on Google’s own announcement of its products.
Why it matters
Opt-in conversation logging creates a new store of user prompts and agent responses, and these may contain personal or confidential data. Teams should assign an owner for `agent_events` tables, set a retention period and record them in the data inventory. The fail-closed skip is tied to existing row-level security, policy tags and masking. Data that is sensitive but not tagged would therefore still be logged, which makes the completeness of classification and tagging a privacy control in its own right.
Status: Available. Token quotas, Looker feedback review and native evaluations are in preview
Sources
Was this useful?
Thanks. What was the problem?
Thank you — this helps us choose what to cover. If you spotted an error, please tell us what it is.
Prepared with AI assistance and reviewed by the editor before publication. Spotted an error? Tell us — see our editorial policy and methodology.
Governance areas
Who should care
- AI governance lead
- Data stewards
- DPO
- Data platform owners