AWS bedrock documentation change
Summary
Added clarification that sensitive information filters only support text output and will not detect PII in tool_use (function call) output parameters via supported APIs
Security assessment
This change documents a limitation of the PII detection feature in Amazon Bedrock Guardrails. While it relates to security functionality (PII detection), it's documenting existing behavior rather than addressing a specific security vulnerability. The change clarifies that PII detection doesn't work with function call outputs, which could impact security if users rely on this feature for those outputs.
Diff
diff --git a/bedrock/latest/userguide/guardrails-sensitive-filters.md b/bedrock/latest/userguide/guardrails-sensitive-filters.md index 75a1e1e84..c069a6354 100644 --- a//bedrock/latest/userguide/guardrails-sensitive-filters.md +++ b//bedrock/latest/userguide/guardrails-sensitive-filters.md @@ -9 +9 @@ Configure sensitive information policy for your guardrail -Amazon Bedrock Guardrails helps detect sensitive information, such as personally identifiable information (PII), in input prompts or model responses using sensitive information filters. This filter is a probabilistic machine learning (ML) based solution that is context-dependent and detects sensitive information based on the context within input prompts or model responses. You can configure by selecting from a set of built-in PIIs offered by Amazon Bedrock Guardrails specific to your use case or organization by defining it along with regular expressions (custom regex) that work based on pattern matching to block or mask PII data. +Amazon Bedrock Guardrails helps detect sensitive information, such as personally identifiable information (PII), in input prompts or model responses using sensitive information filters. This filter supports only text output and will not detect PII information when models respond with tool_use (function call) output parameters via supported APIs. This filter is a probabilistic machine learning (ML) based solution that is context-dependent and detects sensitive information based on the context within input prompts or model responses. You can configure by selecting from a set of built-in PIIs offered by Amazon Bedrock Guardrails specific to your use case or organization by defining it along with regular expressions (custom regex) that work based on pattern matching to block or mask PII data.