Which service filters harmful inputs and outputs for foundation models?
Why A is the answer
Amazon Bedrock Guardrails
Amazon Bedrock Guardrails filters harmful content in both model inputs and model outputs.
In detail
- Guardrails filter harmful inputs and outputs.
- Filters work in both directions.
Use case
An illustrative scenario, not a real incident.
A public chatbot filters harmful input before the model, and harmful output after it.
See it in different ways
flowchart LR A["User input"] --> B["AWS WAF"] B --> C["Input guardrail"] C --> D["Foundation model"] D --> E["Output guardrail"] E --> F["Grounding check"] F --> G["Response"]
sequenceDiagram participant U as User participant I as IAM participant G as Guardrails participant M as Model participant L as CloudTrail U->>I: Is this caller allowed? I->>G: Filter the input G->>M: Safe prompt M-->>G: Output G-->>U: Filtered output I->>L: Records the call
flowchart LR S0["Input filter"] --> S1["Model"] --> S2["Output filter"]
Key point: Guardrails filter inputs and outputs.
Why the other options are wrong
- B. Amazon Macie
- Macie finds sensitive data in S3.
- C. AWS WAF
- WAF filters web requests, not model content.
- D. AWS Key Management Service, by encrypting all of the prompts first
- KMS manages keys.