Bug: creating a guardrail with Content policy sets all filters to high

[dannielvieira-hotmart]

Describe the bug

When creating a Guardrail with a Content policy, setting desired filter to their level results in a Content policy with all filter set to high even though the request didn't specify that.

Regression Issue

• Select this option if this issue appears to be a regression.

Expected Behavior

Expected the guardrail created to set filters that wasnt passed on the request to None.

Current Behavior

Guardrail created with all filters not declared on request set to high

Reproduction Steps

bedrock.create_guardrail(
    name='testing_bug_on_gd_version_1_36',
    contentPolicyConfig={
        'filtersConfig': [
            {
                'type': 'SEXUAL',
                'inputStrength': 'MEDIUM',
                'outputStrength': 'MEDIUM'
            },
            {
                'type': 'VIOLENCE',
                'inputStrength': 'MEDIUM',
                'outputStrength': 'MEDIUM'
            },
        ]
    },
    blockedInputMessaging='string',
    blockedOutputsMessaging='string'
)

--> Visit Bedrock on Console:

Possible Solution

If a filter is not defined, set it as None.

Additional Information/Context

No response

SDK version used

1.36.17

Environment details (OS name and version, etc.)

Mac OS M1 Sonoma 14.7.2

[khushail]

Hi @dannielvieira-hotmart , thanks for reporting this issue.

I tried to repro this with the given code and can see that some configurations like Prompt filters and hate filters were automatically set to High -

This might be an issue on the Service side as I did not see any defaults being mentioned in the API doc -https://docs.aws.amazon.com/bedrock/latest/APIReference/API_GuardrailContentFilterConfig.html

I would file a ticket with the Bedrock team and share the updates.

Thanks

Internal Ticket :P200373051

[dannielvieira-hotmart]

Hi @khushail , thanks!