Kind of. If you have data but don't share it, you can't publish off of it either. And there are grant monitors. If you're ordering the reagents for several times the sequencing you're putting up in the bank, they may well ask questions. Some are more assiduous than others, but do you want to take that chance?
Lessdazed and I are talking about the sharing of utility functions honestly, which is something of a different game than deciding whether to defect on a a cooperative agreement that already has enforcement mechanism in place.
My point was that if it is a Pareto optimum which is going to be implemented then the 'defection' would be in providing a utility function constructed in whatever way would make the calculated pareto optimum closest to your actual utility function. Talking up, understating or outright falsifying your desires in a negotiation is common practice.
Michael Nielsen's new book Reinventing Discovery is invigorating. Here's one passage on how a small group talked an issue through and had a large impact on scientific progress: