The social structures that govern data/sample release aim to safeguard the confidentiality and privacy of cohort research participants (without whom there would be no data or samples) and enable the realization of societal benefit through optimizing the scientific use of those cohorts. Within collaborations involving multiple cohorts and biobanks, however, the local, national, and supranational institutional and legal guidelines for research (which produce a multiplicity of data access governance structures and guidelines) risk impeding the very science that is the raison d'etre of these consortia. We present an ethnographic study, which examined the epistemic and nonepistemic values driving decisions about data access and their consequences in the context of the pilot of an integrated approach to co-analysis of data. We demonstrate how the potential analytic flexibility offered by this approach was lost under contemporary data access governance. We identify three dominant values: protecting the research participant, protecting the study, and protecting the researcher. These values were both supported by and juxtaposed against a “public good” argument, and each was used as a rationale to both promote and inhibit sharing of data. While protection of the research participants was central to access permissions, decisions were also attentive to the desire of researchers to see their efforts in building population biobanks and cohorts realized in the form of scientific outputs. We conclude that systems for governing and enabling data access in large consortia need to (1) protect disclosure of research participant information or identity, (2) ensure the specific expectations of research participants are met, (3) embody systems of review that are transparent and not compromised by the specific interests of one particular group of stakeholders, and (4) facilitate data access procedures that are timely and efficient. Practical solutions are urgently needed. New approaches to data access governance should be trialed (and formally evaluated) with input from and discussion with stakeholders.