Scientists produce vast amounts of data that often are not preserved properly or do not have inventories, placing them at risk. As part of an effort to more fully understand the data-at-risk predicament, researchers who were engaged in the DARI project at UNC's Metadata Research Center surveyed information custodians working in a range of settings. The survey collected information on the data characteristics and preservation plans. Forty-three information custodians completed the survey. The results indicate that atrisk data include a variety of formats, subject areas, and ownership status, as well as compliance with a variety of standards. Although a majority of respondents agree that data preservation is important, they caution that time is the greatest barrier to sharing these data. The study has implications for data rescue and for training information custodians.xperimental and observational data are central to scientific research. With a growing pool of scientific data, researchers have the potential to investigate new questions and use new analytical techniques. To enable this data revolution, the National Science Foundation has funded initiatives to develop cyberinfrastructure for managing, preserving, and sharing scientific data.1 In addition to infrastructure development, there are increasing expectations of open access and new mandates for scientific data management and sharing. For instance, on February 22, 2013, the U.S. Office of Science and Technology Policy released a memo directing federal agencies to increase public access to federal research publications and digital data.2 The memo called for federal agencies with over $100 million in research and development expenditures to develop policies and a plan for access and long-term preservation of government-sponsored research products. Open access presents many challenges for scientists. Scientists are so concerned about data sharing and preservation that the editors of Science recently devoted a special issue to the topic.