I think the general principle for scientific research should be freedom by default, and restriction by exception (incidentally, I would advocate a similar principle for the openness of government/public information, or information within companies, i.e. shared by default, restricted by exception). The other day a scenario that seems a prime candidate for restriction that occurred to me: cellulose metabolism.
Cellulose is the most abundant organic polymer on Earth. It is also relatively difficult to break down (cellulolysis). However, if man were to, through biotechnological research, develop an enzyme that could more efficiently break down cellulose, such an enzyme could pose a significant risk. For example, if this enzyme is engineered into a bacteria (as is common practise in this kind of research) and this bacteria is released into the wild, the competitive advantage to the bacteria at being materially better at breaking down the most abundant food source could result in its rapid proliferation.
Once released, eliminating the bacteria from the wild would be nigh-on impossible due to difficulty in targeting the bacteria and the relatively rapid rate of evolution of bacteria (one possible solution would be nuclear strike on the release zone before the bacteria has had the chance to spread).
The effects of the engineered bacteria could be extensive. Plants existing immune-like mechanisms could potentially be effective, however the increased efficiency of break down would allow such bacteria to breed faster, potentially out-matching these defences. The impact on the world could be apocalyptic.
(The similarity of this thought to the "grey-goo" concept is noted. However, the biological basis for this is more thoroughly understood than self-replicating artificial nano-machines.)