Zero Query Logging (ZQL): what it improves from a privacy perspective

← All articles
Data protection · EN

Zero Query Logging (ZQL): what it improves from a privacy perspective

ZQL is often described as a technical switch, but from a privacy perspective it is better understood as a specific risk‑mitigation measure for web‑grounded AI, not as a general compliance solution.

What ZQL covers

When web grounding is enabled, Copilot may generate a system‑derived search query to retrieve public web content (i.e., via Bing). With ZQL enabled:

1) Bing does not store or retain the generated search queries.
2) No user, tenant, or session identifiers are logged by Bing
3) Queries are processed against a dedicated Copilot web index rather than the standard consumer search pipeline
4) Query visibility and auditing remain on the Copilot / M365 side (e.g. Purview), not in Bing

What ZQL does not change:

1) The web grounding architecture itself remains unchanged (queries still leave the tenant)
2) Bing web search is still outside the Microsoft Data Protection Addendum
3) Microsoft still acts as an independent controller for Bing search processing
4) ZQL does not convert web search into processor‑only processing
5) It does not legitimise the input of personal, confidential, or regulated data into prompts

Put it simply: ZQL is a meaningful improvement in the operational privacy posture of web‑grounded Copilot. It makes certain use cases easier to defend for us (DPOs). But it does not eliminate the need for configuration decisions, user discipline, and documented accountability.

This post has an image on LinkedIn. View on LinkedIn →

Discuss on LinkedIn