I agree overall that paperwork should be minimized. However, a data use agreement is important for data holders to have when writing their IRB (i.e., are they agreeing to release of the data for 1 study or multiple studies, and how long will the data be stored?, who in the future will handle the data if researcher leaves current institution?, etc.).
I think this is important because theoretically data could be obtained for a study by a researcher and then 5 years later they could publish another paper using the data. I am nervous about what could happen if the data holders or generators of the data disagree with how the data are used later on. Also depending on the IRB there could be stipulations about how the data is handled, etc. I believe its better to be upfront about these issues.
If we have a formalized 'Data Use Agreement' as a community then researchers could modify it somewhat along with IRB Protocols, etc. But in that way, everyone knows the restrictions on the data (or the lack thereof) a priori. Long-term this will prevent a lot of problems.
Also many large consortium have such agreements at the outset of joining / obtaining data:
SEER-Medicare has a proposal instruction form and a data use agreement form: http://healthcaredelivery.cancer.gov/seermedicare/obtain/seerdua.docx
that perhaps we could learn from as a community. I don't think we need to be as restrictive. Comments would be very helpful!