Since a flurry of executive orders were issued in early 2025, thousands of datasets and webpages have been removed from federal government websites. The removals have focused on content relating to both topics that have been the subject of executive orders (e.g. gender, structural inequality, climate science, and public health) and content on other topics that uses vocabulary common to research on the topics targeted by the orders.
This guide provides a workflow for researchers needing access to data that has been removed.
Most US federal government datasets are still available on data.gov, if you have not heard specific reports that a dataset has been removed or moved, begin searching by name or topic on data.gov.
If the data is not indexed on Data.gov and you know which government agency produced the information, check their website directly.
The Internet Archive's Wayback Machine indexes many (but not all) .gov webpages. It works best if you have the exact url for the old page - you may be able to find this from cached google search results, citations, etc.
Data Lumos is an ICPSR archive of government data sources that benefit from non-governmental backup.
Tracks current projects to rescue removed government data. Check to see if a dataset you are seeking fits the scope of any projects they recommend.
One of the concerns about the current environment of data removal is that datasets hosted directly on government websites have verified provenance, and copies of this data access may lack assurances that they have not been edited. If you are planning to assist in data archiving efforts, follow MIT's checklist for federal government backups. This checklist provides simple steps you can take to document the provenance of the dataset you're saving and help users confirm that it has not been edited.
The checklist can also be used to discover if datasets still on federal websites or uploaded to data repositories have been edited. If a Wayback Machine copy of the data exists, compare checksum and other metadata to determine if a copy in a repository or on a federal website matches the original data.