Plants are replete with thousands of proteins and small molecules, many of which are species-specific, poisonous or dangerous. Over time humans have learned to avoid dangerous plants or inactivate many toxic components in food plants, but there is still room for ameliorating food crops (and plants in general) in terms of their allergens and toxins content, especially in their edible parts. Inactivation at the genetic rather than physical or chemical level has many advantages and classical genetic approaches have resulted in significant reduction of toxin content. The capacity, offered by genetic engineering, of turning off (inactivating) specific genes has opened up the possibility of altering the plant content in a far more precise manner than previously available. Different levels of intervention (genes coding for toxins/allergens or for enzymes, transporters or regulators involved in their metabolism) are possible and there are several tools for inactivating genes, both direct (using chemical and physical mutagens, insertion of transposons and other genetic elements) and indirect (antisense RNA, RNA interference, microRNA, eventually leading to gene silencing). Each level/strategy has specific advantages and disadvantages (speed, costs, selectivity, stability, reversibility, frequency of desired genotype and regulatory regime). Paradigmatic examples from classical and transgenic approaches are discussed to emphasize the need to revise the present regulatory process. Reducing the content of natural toxins is a trade-off process: the lesser the content of natural toxins, the higher the susceptibility of a plant to pests and therefore the stronger the need to protect plants. As a consequence, more specific pesticides like Bt are needed to substitute for general pesticides.