What is zero data retention (ZDR)?

Zero data retention (ZDR) is the claim that an AI endpoint does not keep the prompt, the completion, or other residue of a request after the call finishes.

Microsoft Azure OpenAI / Foundry, modified abuse monitoring zero data retention: terms and exceptions

Q: What is the ZDR score for Microsoft Azure OpenAI / Foundry, modified abuse monitoring?

On zdr.dev the published score is ZDR 1.0 (s,f,h,r) · B. No OpenAI access; abuse monitoring w/ flagged-sample human review (automated LLM review default); store duration removed from canonical page. Automated review continues under MAM; severe abuse can force monitoring back on; stateful features persist; stricter criteria for advanced models.

ZDR 1.0 (s,f,h,r) · B

Short answer

No OpenAI access; abuse monitoring w/ flagged-sample human review (automated LLM review default); store duration removed from canonical page. Automated review continues under MAM; severe abuse can force monitoring back on; stateful features persist; stricter criteria for advanced models.

Scope and evidence

Provider	Microsoft
Service	Azure OpenAI / Foundry, modified abuse monitoring
Lane	Enterprise
Origin	US United States
License	proprietary
Hosting	Provider-hosted
Score	ZDR 1.0 (s,f,h,r) · B Contractual ZDR. Paper ZDR: no content retention by agreement, with typed carve-outs in force
Evidence	Public documentation reviewed (B: published documentation)
Source-review date	2026-08-24 (registry)

Other lanes for Microsoft

ZDR 1.0 (s,f,h,r) · Bmicrosoft/azure-openai · foundryEnterprise

Structured breakdown

Default posture: no OpenAI access; abuse monitoring w/ flagged-sample human review (automated LLM review default); store duration removed from canonical page
Exceptions and carve-outs: automated review continues under MAM; severe abuse can force monitoring back on; stateful features persist; stricter criteria for advanced models

Approval and setup path

Limited Access form, account-managed customers; verify via ContentLogging:false

Flags on this record

s: safety-flag retention beyond base window
f: stateful-feature carve-outs (files, batch, threads, caching, grounding)
h: human review possible
r: revocable: provider reserves unilateral downgrade

Practical questions for your agreement

Does your signed agreement cover this exact service, model set, and feature path?
Which retention, logging, and human-review exceptions still apply after any ZDR election?
What configuration evidence (flags, project settings, store defaults) confirms the posture you expect?
Do your customer commitments match what the reviewed evidence supports?