Having identified a list of possible harms that might occur as a result of using or releasing genetically engineered organisms, the next question is how likely are any of these to occur? Like the original "brainstorming" of potential harms, the answer to this question depends greatly on how well the organisms and their interaction in the environment are understood.

Risks must be assessed case by case as new applications of genetic engineering are introduced. In some circumstances, it is possible to assess risks with great confidence. For example, it is vanishingly unlikely that genetically engineered palm trees will thrive in the Arctic regardless of what genes have been added. But for many potential harms, the answers are far less certain.

Risk assessments can be complicated. Because even rigorous assessments involve numerous assumptions and judgment calls, they are often controversial when they are used to support particular government decisions. For example, the approval of the first genetically engineered squash by the United States Department of Agriculture involved a controversial risk assessment.

Under the current US regulatory framework for biotechnology, some sort of risk assessment is routinely produced before decisions to allow commercialization of products under the Federal Plant Pest Act; the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA); and the Toxic Substances Control Act (TSCA).

In the case of the Plant Pest Act, risk assessments are done according to the procedure specified by the National Environmental Policy Act (NEPA). Under NEPA, risk assessments could lead to full-blown environmental impact statements, but so far all evaluations of engineered agricultural organisms have led to the legal conclusion that no environmental impact statement is needed.

For the most part, risk assessments are done by scientists and policymakers in the relevant agencies (USDA or EPA) with information provided by the companies seeking the approvals. The public often has a brief opportunity to review and comment on the risk assessments.

There is no standard set of questions that risk assessments must answer because of the great range of potential impacts of biotechnology products. A risk assessment for a microbial pesticide, for example, would be substantially different from a risk assessment for genetically engineered salmon. Like all efforts at risk evaluation, risk assessments done for regulation depend on the base of scientific knowledge for generation of list of possible harms to be assessed.

Finally, there are so-called projectile methods that use metal slivers to deliver the genetic material to the interior of the cell. The small slivers (much smaller than the diameter of the target cell) are coated with genetic material. One projectile method, called bioballistics, propels the coated slivers into the cell using a shot gun. A perforated metal plate stops the shell cartridge, but allows the slivers to pass through and into the living cells on the other side. Once in the cell, the genetic material is transported to the nucleus where it is incorporated among the host genes.
Contrary to the arguments made by some proponents, genetic engineering is far from being a minor extension of existing breeding technologies. It is a radically new technology for altering the traits of living organisms by inserting genetic material that has been manipulated by artificial means. Because of this, genetic engineering may one day encompass the routine addition of novel genes that have been wholly synthesized in the laboratory.

Novel organisms bring novel risks, however, as well as the desired benefits. These risks must be carefully assessed to make sure that all effects-both desired and unintended-are benign. UCS advocates caution, examination of alternatives, and careful case-by-case evaluation of genetic enginering applications within an overall framework that seeks to move agricultural systems of food production toward sustainability.