Redactive

Redact keywords and text patterns, redact SSNs, account numbers, and more. Redaction, by definition, means removing certain types of information from documents. In the context of United States government agency documents, redaction refers to the process of removing classified information from a document prior to its publication. For attorneys, redacting is a very important procedure for protecting confidential information. In case of discovery a document must be distributed containing sensitive information, all non-discoverable confidential material needs to be redacted from that document first. AutoRedact is a "true redacting" solution that deletes sensitive text, annotations, text form fields, vector graphics, and image information from PDF documents making them completely safe for distribution. The software contains many powerful tools that increase productivity and provide automated document redacting and multi-user redacting workflows.

Amazon Bedrock Guardrails is a configurable safeguard system designed to enhance the safety and compliance of generative AI applications built on Amazon Bedrock. It enables developers to implement customized safety, privacy, and truthfulness controls across various foundation models, including those hosted within Amazon Bedrock, fine-tuned models, and self-hosted models. Guardrails provide a consistent approach to enforcing responsible AI policies by evaluating both user inputs and model responses based on defined policies. These policies include content filters for harmful text and image content, denial of specific topics, word filters for undesirable terms, sensitive information filters to redact personally identifiable information, and contextual grounding checks to detect and filter hallucinations in model responses.