[FEATURE] Amazon Comprehend Solution for PII Detection and Redaction in Bedrock

[liamschn]

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

Is your feature request related to a problem? Please describe

Generative AI models and associated data can potentially contain sensitive information, personally identifiable information (PII), or other regulated data. Without proper controls in place, there is a risk of inadvertent disclosure or leakage of this sensitive data, which can lead to compliance violations and legal issues.

Describe the solution you'd like

Develop an Amazon Comprehend solution to automatically detect and redact PII and other sensitive information from the data ingestion pipeline and model outputs in the Bedrock environment. This solution should leverage Amazon Comprehend's natural language processing capabilities to identify and mask sensitive data before it is processed or generated by the AI models.

Describe alternatives you've considered

An alternative approach is to manually review and sanitize data before ingesting it into the Bedrock environment. However, this is a time-consuming and error-prone process, especially with large volumes of data.

Additional context

The solution should incorporate appropriate security controls, such as encryption at rest and in transit, secure networking, and access controls. It should also include mechanisms to detect and redact PII and other sensitive information from the Bedrock data pipeline.