Until recently, most U.S. Census Bureau data products used traditional statistical disclosure limitation (SDL) methods such as cell or item suppression, data swapping, input noise injection, and censoring to protect respondents’ confidentiality. In response to developments in mathematics and computer science since 2003 that have significantly increased the risk of reconstruction and re-identification attacks, the Census Bureau is developing formally private SDL methods to protect its data products. These methods provide mathematically provable protection for respondent data and allow policy makers to manage the tradeoff between data accuracy and privacy protection—something previously done by technical staff. The first Census Bureau product to use formal methods for privacy protection was OnTheMap, a web-based mapping and reporting application that shows where workers are employed and where they live. Recent research for OnTheMap is implementing formal privacy guarantees for businesses to complement the existing formal protections for individuals. Research is underway to improve the disclosure limitation methods for the 2020 Census of Population and Housing, the American Community Survey, and the 2022 Economic Census. For each of these programs, we are developing new state-of-the-art privacy protection approaches based on formal mechanisms that have been vetted by the scientific community. There are many challenges in adopting formally private algorithms to datasets with high dimensionality and the attendant sparsity. In addition to formally private methods that allow senior executives to set the privacy-loss budget, our implementations will feature adjustable “sliders” for allocating the privacy-loss budget among related statistical products. The Census Bureau is implementing the settings for the privacy-loss budget and these sliders based on the decisions of the Census Bureau’s Data Stewardship Executive Policy Committee.