For papers you would just do what NIH requires, which is "upon acceptance for publication." Upon further reflection and discussion data would probably be more reasonably published in an open format no later than 1 year after the grant ended. That would allow multi-year studies a little more time to collect all of their data and write it up.
I was just commenting on the existing open access policy (edit: I noticed you opened the subthread by prescribing, sorry for derailing).
Open science/sharing the data can work as a voluntary movement, it has not been captured by private interests, but I'm really interested in seeing Open Access pan out before that. The current policy, if I'm reading it right, is saying that the manuscripts must be submitted to PubMed Central, but PMC or the researcher (whoever decides to trigger it first I imagine) only has an obligation to release them after twelve months. If PubMed Central doesn't release the research instantly and the researchers aren't aware of their rights or are subjet to some pressure/contradicting (possibly illegal) agreements, there's a loophole in the policy as written.
