Safety, RAG & search
Amazon Bedrock Guardrails vs Foundry guardrails (content filters) / Azure AI Content Safety
The Azure equivalent of Amazon Bedrock Guardrails is Foundry guardrails, which are powered by Azure AI Content Safety.
Content Safety is also available as a standalone API, like Bedrock's ApplyGuardrail.
Side-by-side comparison
| Aspect | Amazon Web Services | Microsoft Azure |
|---|---|---|
| Service | Amazon Bedrock Guardrails | Foundry guardrails (content filters) Azure AI Content Safety |
| Lifecycle | Generally available | Agent guardrails, tool-call/tool-response intervention points, and PII risk are in preview. Generally available |
| What it is | Content filters (hate, insults, sexual, violence, misconduct, prompt attack), denied topics, word filters, PII filters, contextual grounding checks, and Automated Reasoning checks. Also usable standalone through the ApplyGuardrail API. | Named collections of controls applied to models and agents at user input, tool call, tool response, and output, with severity-based blocking. Standalone text and image moderation APIs, Prompt Shields (jailbreak and indirect-attack detection), groundedness detection (preview), protected-material detection, custom categories (preview), and agent task-adherence detection. |
| Pricing model | Charged per text unit processed by each policy type (content, denied topics, PII, grounding, Automated Reasoning). Word filters are free of charge. | Content filtering on Foundry deployments is included with model usage. Standalone Content Safety APIs are billed per text record and per image, with a free tier. |
| Docs | Amazon Bedrock Guardrails docs | Foundry guardrails (content filters) docs Azure AI Content Safety docs |
| Pricing page | Amazon Bedrock Guardrails pricing | Foundry guardrails (content filters) pricing Azure AI Content Safety pricing |
| Exams | AIF-C01 Generative AI Developer – Professional | AI-901 AI-103 |
Key differences
- Content categories differ. Bedrock filters Hate, Insults, Sexual, Violence, Misconduct, and Prompt Attack. Azure classifies Hate, Sexual, Self-harm, and Violence by severity, with Prompt Shields for jailbreak and indirect attacks.
- Only on Bedrock, Automated Reasoning checks (logic-rule validation of responses), plus denied topics and word filters (Azure uses blocklists and custom categories).
- Only on Azure, protected-material detection (known text and code) and task-adherence detection for agent tool use. Guardrails can also run at tool-call and tool-response intervention points for agents (preview).
- Hallucination checks are on both sides. Bedrock has contextual grounding checks, and Azure has groundedness detection (preview).
- Turning output filtering off on Azure-sold models requires Microsoft approval (Limited Access, modified guardrails).
- Bedrock offers cross-account enforcement of guardrails and Classic vs Standard safeguard tiers.
Pricing summary
Pricing models only. No prices are quoted because they change by region and over time.
- AWS
- Charged per text unit processed by each policy type (content, denied topics, PII, grounding, Automated Reasoning). Word filters are free of charge. Amazon Bedrock Guardrails pricing →
- Azure
- Content filtering on Foundry deployments is included with model usage. Standalone Content Safety APIs are billed per text record and per image, with a free tier. Foundry guardrails (content filters) pricing → Azure AI Content Safety pricing →
Certification exam mapping
| Exam | Provider | Level | Status |
|---|---|---|---|
| AWS Certified AI Practitioner AIF-C01 | AWS | Foundational | current Italian and German versions retire after Oct 15, 2026 (per AWS cert page). |
| AWS Certified Generative AI Developer – Professional | AWS | Professional | current Exam code not confirmed from the official page at build time; shown by name only. |
| Microsoft Certified – Azure AI Fundamentals AI-901 | Azure | Fundamentals | current AI-901 replaces AI-900 for this certification (English version updated Apr 15, 2026). Covers Microsoft Foundry. |
| Microsoft Certified – Azure AI Apps and Agents Developer Associate AI-103 | Azure | Associate | current Generative AI and agentic solutions with Microsoft Foundry using Python. Successor in scope to the retired AI-102. |
Relevance is based on each exam's published scope, not a line-by-line objective match. Check the official study guide.
Verification notes
- "Word filters free" reflects Bedrock pricing-page structure. Confirm current pricing.
FAQ
What is the Azure equivalent of Amazon Bedrock Guardrails?
The Azure equivalent of Amazon Bedrock Guardrails is Foundry guardrails, which are powered by Azure AI Content Safety. Content Safety is also available as a standalone API, like Bedrock's ApplyGuardrail.
How does pricing compare between Amazon Bedrock Guardrails and Foundry guardrails (content filters) and Azure AI Content Safety?
AWS: Charged per text unit processed by each policy type (content, denied topics, PII, grounding, Automated Reasoning). Word filters are free of charge. Azure: Content filtering on Foundry deployments is included with model usage. Standalone Content Safety APIs are billed per text record and per image, with a free tier. Check the official pricing pages for current rates.
Which certification exams cover this topic?
AWS Certified AI Practitioner (AIF-C01); AWS Certified Generative AI Developer – Professional; Microsoft Certified – Azure AI Fundamentals (AI-901); Microsoft Certified – Azure AI Apps and Agents Developer Associate (AI-103).
Does Azure have an equivalent to the Bedrock ApplyGuardrail API?
Yes. Azure AI Content Safety exposes text and image moderation, Prompt Shields, groundedness, and protected-material detection as standalone APIs you can call on any content.