Configurable safeguard service inside Amazon Bedrock that evaluates user inputs and model responses against content filters, denied topics, sensitive information filters and word filters, including a prompt attack category. It can be applied at inference or via a standalone API, but does not authorize agent tool calls.
commercial · generally available · Research snapshot 2026-09-06
Show me a Bedrock guardrail blocking a prompt attack and masking PII for a Bedrock Agent, then show the same guardrail invoked through ApplyGuardrail for a non-Bedrock model.
Capabilities and evidence
Support labels reflect the supplied research. Documentation and vendor claims are not independent product tests. “Not established” means the researcher did not find support; it does not prove a capability is absent.
Documented by provider
Guardrails evaluate both user inputs and model responses against configured policies including content filters with a Prompt Attack category, denied topics, sensitive information filters and word filters, returning blocked messaging or masked content on intervention.
Limit: Documentation does not describe detection of tool misuse, agent action authorization or multi-step agent behaviour analysis.
Guardrails can be applied during foundation model inference by specifying a guardrail ID and version, or invoked independently through the ApplyGuardrail API without calling a model, and are supported with Bedrock Agents and Knowledge Bases.
Limit: Cross-account and cross-region behaviour is documented separately and was not verified here.