- Newest
- Most votes
- Most comments
I understand your concern about the Bedrock Guardrail prompt attack filter not changing as expected. Based on the information provided, it seems you're experiencing an issue where the filter strength for the PROMPT_ATTACK type is defaulting to HIGH, regardless of your attempts to set it to different values.
First, it's important to note that Amazon Bedrock Guardrails is a feature that allows you to configure content filters with adjustable thresholds across different content categories, including prompt attacks. The ability to customize these filters is a key feature of the service.
However, your experience of being unable to change the filter strength, particularly for the prompt attack filter, is not the expected behavior. Normally, you should be able to modify these settings through the console or API.
The fact that this worked differently a couple of months ago suggests there might have been a recent change or update to the service. Amazon Bedrock and its features, including Guardrails, are continuously evolving, and some aspects are still in preview.
Given that you're unable to change the setting through multiple methods (update, console changes, and recreating), this could potentially be a bug or an intentional change in how the prompt attack filter operates.
To address this issue, I would recommend the following steps:
- Double-check that you have the necessary permissions to modify these settings.
- Ensure you're using the most up-to-date version of any SDKs or CLI tools you might be using.
- Contact AWS Support for clarification. They will be able to provide the most current information about any changes to the service or confirm if this is indeed a bug that needs to be addressed.
Remember, Amazon Bedrock and its features are subject to updates and changes, especially while in preview. It's always a good idea to stay informed about the latest developments and best practices for using these services.
Sources
Generative AI Data Governance - Amazon Bedrock Guardrails - AWS
GuardrailContentFilter - Amazon Bedrock
answered 2 years ago
Relevant content
asked 2 years ago
asked 2 years ago
- AWS OFFICIALUpdated a year ago
- AWS OFFICIALUpdated 5 months ago
