Core Features and Prototypes

The AI-enabled National Secure Data Service (NSDS) is a flagship initiative of the Federal Statistical System. Below are brief descriptions of select NSDS core features aligned to the challenges the tool or service is addressing.

AI Data Harmonizer Toolkit

This toolkit offers a suite of five tools, addressing key challenges to meaningfully accelerate analyst capacity:

Challenge	Tool
Lack of standardized and complete metadata for tabular datasets.	The Metadata Extractor uses AI to extract metadata from tabular data files.
Difficulty analyzing free-text fields in survey and administrative data at scale.	The Free Text Encoder uses AI to enable users to cluster or categorize free-text fields.
Challenge of extracting structured information from semi-structured or unstructured documents, such as regulatory filings and public reports.	The Tabular Data Extractor uses AI to enable users to extract structured answers and tabular data from semi-structured documents (e.g., PDFs, TXT files).
Difficulty of standardizing tabular data from multiple sources that represent similar concepts but differ in structure, naming conventions, or formatting.	The Data Harmonizer uses AI to harmonize multiple tabular datasets by extracting schemas, defining or generating a target schema, grouping similar fields, and matching values across files.
The need for scalable, transparent, and accurate methods to handle missing data in administrative and survey datasets.	The Missing Data Assistant uses a browser-based interface to offer sophisticated imputation methods.

AI Ready Toolkit

Challenge	Tool
Many Federal statistical agency websites are not formatted in a way that can be consistently optimized by AI tools.	The AI-Ready Evaluator assesses whether a website’s content is AI-ready, what percent is AI-ready, and whether the URL can be run using the transformation tool.
Many Federal statistical agency websites are not formatted in a way that can be consistently optimized by AI tools.	The AI Ready Transformation Tool determines tagging required to render website content AI-ready and provides the code that can be utilized by a systems engineer. It is a companion tool for the AI Readiness Evaluator.

AI State Data Explorer

Challenge	Tool
Many state, local, and territorial governments do not have the infrastructure or capacity to take full advantage of the administrative data under their control.	This project provides state governments with a platform that includes a conversational AI interface that supports enhanced evidence building.

Navigator

Government agencies collect and produce an abundance of high-quality data, which fuels critical research and sound policy. However, it is challenging to find relevant data or data tools across multiple different agency websites. Agencies, moreover, have no central resource that shows how their data is being used, and data users have no central space to share questions. The Data Navigator addresses these gaps, serving as an intelligent front door to the nation’s data. The Navigator has three main components:

Challenge	Tool
The Federal statistical system is difficult even for expert data users to navigate, and there is currently no central location for data users to ask open-ended questions.	The Navigator AI Chatbot answers natural language questions with information sourced directly from Federal statistical agency websites. Users with more complex questions can request assistance from the human Subject Matter Expert Network.
Currently, there is no shared resource for federal agencies to track the usage of their data assets or for data users to find others working on the same topics or datasets.	The Data Usage Platform tracks federal data usage across the publication and internet ecosystem, with a dashboard interface for agencies.
Data users working with government statistical data are often siloed, without the ability to ask questions based on shared research topics.	The Community Forum offers a curated space for data users to ask questions based on shared research topics.

Privacy-enhancing Technologies

Data users and providers face multiple challenges in securely accessing and provisioning data for evidence-building. Privacy-enhancing technologies (PETs) offer solutions to address these barriers. Many agencies are exploring the use of PETs to securely expand access and linkage opportunities. The NSDS has conducted and continues to test several pilots using PETs to link data and support tiered access.

Challenge	Tool
Data providers face multiple challenges to securely link or share data across agencies.	The NSDS has conducted two successful pilots with privacy-preserving record linkage and can point users to open-source code and commercial tools for future use cases. In addition, NSDS can provide lessons learned, support for data sharing agreements, and guidance on use of these tools. An NSDS pilot is currently testing the use of a Secure Multiparty Computation tool to link data from a government agency and private sector data. The tool will be available through the NSDS secure compute environment.
Privacy protection may limit access to data and may require data users to request access to confidential microdata, an access mode that is resource and time-intensive for agencies.	NSDS pilot projects have successfully developed synthetic datasets and can direct users to open-source code for synthetic data generation to support tiered access.

Challenge

Tool

Data providers face multiple challenges to securely link or share data across agencies.

The NSDS has conducted two successful pilots with privacy-preserving record linkage and can point users to open-source code and commercial tools for future use cases. In addition, NSDS can provide lessons learned, support for data sharing agreements, and guidance on use of these tools.

An NSDS pilot is currently testing the use of a Secure Multiparty Computation tool to link data from a government agency and private sector data. The tool will be available through the NSDS secure compute environment.

Privacy protection may limit access to data and may require data users to request access to confidential microdata, an access mode that is resource and time-intensive for agencies.