Amazon Bedrock Guardrails

Amazon Bedrock Guardrails is a cloud service for applying controls to model interactions. It is best suited to teams that need to choose among content, information and prompt-attack protections. Its appeal is the range of individually priced checks; the trade-off is a paid model with no free plan.

Overview

Guardrails combines content filters, denied topics, sensitive-information filters, word filters and contextual grounding checks with prompt-injection defense and output guardrails. It can address risks in both prompts and responses, including PII and secrets. That breadth suits teams with defined protection needs, but the separate charges make it important to select controls deliberately rather than assume the full set comes as one package.

Key features

Content and topic controls

Text content filters and denied-topic checks are separate controls, each priced at $0.15 a month on Standard and Classic tiers. Image content filters cost nothing. The split lets teams distinguish between text and image filtering, while the paid text options make the cost depend on which controls an application needs.

Sensitive information and grounding

Sensitive-information filters cost $0.10 a month and support PII redaction and secrets detection. Regular-expression sensitive-information filters cost nothing, as do word filters. Contextual grounding checks cost $0.10 a month, making them another separately priced choice for teams that need this check alongside content or information controls.

Prompt-attack checks

Prompt-injection defense is relevant to applications that accept user prompts, while output guardrails extend the control surface to model responses. InvokeGuardrailChecks offers text-only content filters for $0.07 a month and text-only prompt-attack checks for $0.08 a month; its sensitive-information filter checks cost $0.10 a month. These distinct prices give teams options for specific checks, but text-only coverage on two of them may not suit workloads needing those checks on other content types.

Pricing

Amazon Bedrock Guardrails is paid and has no free plan. Its controls are priced separately: Content filters — text cost $0.15 a month on Standard and Classic tiers; Content filters — image cost nothing; Denied topics cost $0.15 a month on Standard and Classic tiers; Sensitive information filters cost $0.10 a month; Sensitive information filters — regular expression and Word filters each cost nothing; and Contextual grounding checks cost $0.10 a month. InvokeGuardrailChecks — content filters cost $0.07 a month for text content only, InvokeGuardrailChecks — prompt attack cost $0.08 a month for text content only, and InvokeGuardrailChecks — sensitive information filters cost $0.10 a month.

The zero-priced controls do not change the fact that the product has no free plan. Teams comparing costs should map their required protections to the individual control prices; there is no single plan price to weigh against an all-inclusive feature set.

Platforms

Guardrails is a cloud service with web access. That fits teams seeking cloud deployment, but it is not positioned here as a self-hosted option.

Who it's for

Guardrails fits teams building model interactions that need specific input and output safeguards, particularly prompt-injection defense, PII redaction, secrets detection or contextual grounding checks. It is a less compelling choice for readers who need a free plan or want one bundled price for multiple protections.

Pros and cons

• Pros: A broad set of controls covers content, denied topics, sensitive information, words, grounding, prompt attacks and outputs.
• Pros: Separate control pricing lets teams select protections rather than pay a single stated price for every check.
• Cons: There is no free plan, so teams cannot start with a free tier.
• Cons: Multiple checks have separate charges, which makes the total depend on the selected protections.
• Cons: Two InvokeGuardrailChecks options are text-only, which limits their fit for workloads needing those checks on other content types.

Alternatives

PromptGuard is worth considering for a freemium alternative: its free tier includes 20,000 scans a month, one API key, one project and 24-hour log retention, while Team costs $19 a month. Enkrypt AI Guardrails offers a free Explore tier with 500 starting credits, 50 credits per month, seven-day data retention and community support; its Launch plan costs $134 a year.

Fiddler Guardrails has a free tier for real-time guardrails against harmful exposure, hallucinations, toxicity, PII/PHI, prompt injection and jailbreak attempts. Openlayer Guardrails may suit a reader looking for a free starting tier with one member, five projects, one inference pipeline per project and 20,000 inferences per month.

OpenSecureAI Prompt Firewall has a free plan with in-browser tools and npm packages, 2,000 API requests per month, and a 30-requests-per-minute-per-key limit. HiddenLayer AI Runtime Security is another paid option. OpenAI Guardrails is a free alternative for API, self-hosted and web platforms. Oracle Mobile Authenticator is a free option for Android, iOS and Windows.

For broader comparisons, browse AI Guardrail Software or LLM Security Tools.

Verdict

Choose Amazon Bedrock Guardrails when your team needs a configurable mix of prompt, information and output protections and is prepared to pay for selected checks. The range of controls is its main strength; teams that need a free starting tier or prefer a single bundled price should look elsewhere.