Azure AI Content Safety (Prompt Shields) by Microsoft Azure

# Fix list: Azure AI Content Safety (Prompt Shields)

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/azure-ai-content-safety, the October 2026 research run, assessed 1 October 2026. Grade C, 60.9 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on Azure AI Content Safety (Prompt Shields): work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Payments & pricing, 15 out of 100, up to 10.6 more on the total

Why it scored 15: No x402, MPP or L402 (0). Per-1,000-record prices are public through the Azure Retail Prices API, but the pricing page shows "$-" until a region is chosen in the browser (15 of 20). F0 gives 5,000 free text records and 5,000 images a month, but it needs an Azure subscription and an Azure account needs a card (0). A person signs up and creates the resource in a browser (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 2. Reliability, 55 out of 100, up to 9 more on the total

Why it scored 55: Azure status page with a post-incident review history (20). Three reviews in the last 90 days touch the service's neighbourhood, a West US network incident on 23 July, intermittent failures and latency across Azure OpenAI, Foundry and Cognitive Services in Sweden Central on 29 September (10:03 to 15:58 UTC), and a multi-region connectivity incident on 30 September (about 5 hours 45 minutes). Content Safety isn't named, but it's a Cognitive Services resource, so we count one major (10). Rate limits per feature and tier in the overview, 5 a second on F0 and 1,000 per 10 seconds on S0 for text, images and Prompt Shields (15). No 429 or backoff guidance in the Content Safety docs or the Shield Prompt reference (0). Microsoft's Online Services SLA is a downloadable document (1 October 2026 edition) that we couldn't read, so we couldn't confirm Content Safety is covered (0). Text analysis, image analysis and Prompt Shields are GA on api-version 2024-09-01 (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 3. Schema & documentation, 69 out of 100, up to 5 more on the total

Why it scored 69: Public OpenAPI (Swagger) documents in Azure/azure-rest-api-specs, with stable 2024-09-01 and previews up to 2026-09-01-preview, error schemas and examples on all 15 operations (25). No llms.txt at learn.microsoft.com (0). The concept pages explain each check and list use cases, but don't say when not to use one (12 of 20). Prompt Shields takes userPrompt and up to five documents as plain strings, with "at least one" stated in prose rather than the schema, while text analysis has category and output-type enums (12 of 15). Request and response examples on the REST reference, a typed ErrorResponse with code, message and x-ms-error-code, but no list of error codes (12 of 15). api-version on every call and a What's New page, but its last entry is November 2025 while the 2026-07-01-preview and 2026-09-01-preview versions appeared in the spec repository (8 of 15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 4. Maintenance & community, 45 out of 100, up to 4.8 more on the total

Why it scored 45: A 2026-09-01-preview API version is in Microsoft's public spec repository (30). One dated API version in the last 90 days, the 2026-07-01-preview falling just outside (0). The What's New page hasn't recorded anything since the Task Adherence preview in November 2025, though Microsoft Q&A and support answer (5 of 15). The official SDKs lag the service, Python 1.0.0 from December 2023 and JavaScript 1.0.1 from January 2025, neither with a Prompt Shields helper (5 of 15). Packages unchanged for 21 and 9 months (5 of 10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## 5. Security & auth, 74 out of 100, up to 4.6 more on the total

Why it scored 74: Microsoft Entra ID tokens with RBAC, or two resource keys that can be regenerated in turn (30). RBAC can limit a caller, but a resource key also reaches the blocklist write and delete operations, with no confirmation step (15 of 20). Prompt Shields detects user-prompt and document attacks, and the docs describe Spotlighting for third-party content, in preview (15). We didn't find per-call logging documented for Content Safety (0). Coordinated disclosure policy and MSRC bounty programmes linked from security.txt, but microsoft.com's security.txt expired on 23 September 2026 (4 of 5), bounty (5), SOC or ISO coverage for Content Safety by name not confirmed (0), MSRC publishes advisories (5), so 14 of 20.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 6. Agent ergonomics, 78 out of 100, up to 3.6 more on the total

Why it scored 78: Prompt Shields answers one boolean per prompt and per document, and text analysis returns four or eight severity levels by choice (20 of 25). Categories, output type and haltOnBlocklistHit are set per request (20). Errors carry a code and message in a standard Azure shape, but the codes aren't documented per operation (15 of 20). Checks have no side effects, but there's no retry guidance (15 of 20). Few required fields, but the Python SDK is 1.0.0 from 12 December 2023 with no Prompt Shields method, and the JavaScript package is a generic REST client, 1.0.1 from January 2025 (8 of 15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 7. Transparency & trust, 83 out of 100, up to 1.5 more on the total

Made of editorial 70, provenance 95.

Why it scored 83: Closed service under Microsoft's product terms (15). The FAQ and the data-privacy page agree that inputs aren't stored, aren't used for training and stay in the resource's region, and that only customer blocklists are kept, encrypted, in that region (25 of 30). The 2025-03-01 retirement of older API versions was announced in October 2024 with the date (15 of 20). Processing stays in the resource's region, and regions are listed per feature, but we didn't check a subprocessor list (15 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- security.txt: published but past its Expires date (5 of 10)

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- Whether Microsoft's Online Services SLA names Azure AI Content Safety, and at what percentage.
- Whether Content Safety supports diagnostic request logging per call, which we didn't find in its docs.
- Whether the 2026-07-01-preview and 2026-09-01-preview versions are live in every region, since What's New doesn't mention them.
- Whether Content Safety is named in Microsoft's SOC 2 and ISO 27001 scope.

## Weaknesses

- Needs an Azure subscription with a card, a resource and a region that has the feature, before the first call
- Python SDK is 1.0.0 from December 2023 and has no Prompt Shields method
- No retry or 429 guidance in the Content Safety docs
- What's New hasn't been updated since November 2025, while 2026 preview API versions appeared in the spec repository
- 10,000 characters per request, documents included, so long tool results have to be chunked

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Send retrieved pages and tool results in the documents array of shieldPrompt, not in userPrompt, so document attacks are reported separately
- Call text:shieldPrompt over REST with api-version=2024-09-01. The Python SDK 1.0.0 has no method for it
- Keep each request under 10,000 characters across prompt and documents, and split long tool results
- Create the resource in a region that lists Prompt Shields, since not every region has it
- On F0 you get 5 requests a second. Queue checks or move to S0 before load testing

## What the review panel asked for

- Add a Prompt Shields method to the Python SDK
- List the error codes per operation
- a data-plane key without write access
- per-call request logging

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.