I recently came across a striking example of this. In 1983, Richard Louttit of the National Science Foundation, who had helped craft the list of exemptions encoded in 45 CFR 46.101, explained them as follows:
Much research of minimal risk was exempted from IRB review in order to reduce the IRB workload so that research involving ethical questions could get more than cursory review. But some institutions have decided that the IRB, or its chairperson, must review proposals to decide if they are exempt from review. If this seems contradictory, it is. And this was not envisioned by the staff group which worked out the exemptions.
[Richard T. Louttit, "Government Regulations: Do They Facilitate or Hinder Social and Behavioral Research?," in Joan E. Sieber, ed., NIH Readings on the Protection of Human Subjects in Behavioral and Social Science Research: Conference Proceedings and Background Papers (Frederick, Md: University Publications of America, 1984), 179.]
Yet in 1995, the Office for Protection from Research Risks adopted just that contradictory position:
Institutions should have a clear policy in place on who shall determine what research is exempt under .46.101(b). Those persons who have authority to make a determination of what research is exempt are expected to be well-acquainted with interpretation of the regulations and the exemptions. In addition, the institution should be prepared to reinforce and review, as necessary, the method of determining what is exempt. OPRR advises that investigators should not have the authority to make an independent determination that research involving human subjects is exempt and should be cautioned to check with the IRB or other designated authorities concerning the status of proposed research or changes in ongoing research.
(OPRR Reports, 95-02) [21 July 2014 updated link.]
The regime in place today is far more intrusive than the one worked out in 1981. Changes in the regulations themselves are part of the problem, but so are radical reinterpretations like the one above.