This paper analyzes economic issues that arise in devising a credible and enforceable system of identity preservation and labeling for genetically modified (GM) and non-GM products. The model represents three stages in the supply chain: farm production, marketing handlers, and final users. The possibility of accidental co-mingling of non-GM products is modeled at the marketing stage. Regulation takes the form of a threshold level of purity for non-GM products, a probability of government testing to verify compliance with the threshold level, and a fine for violators. Uncertainty is modeled explicitly, such that would-be suppliers of non-GM products always face some risk of failing the test and incurring a fine. The paper also presents a novel demand specification for differentiated GM and non-GM products that is particularly useful in our stochastic framework. The results emphasize the role and impact of an uncertain testing technology, a critical feature in this setting. We also highlight the somewhat nonstandard trade-off between frequency of testing and the size of the fine that applies here. Because testing can prevent mislabeled product from reaching the consumer, it can provide a direct welfare benefit. An equilibrium that includes production of the non-GM good may only be supportable with sufficiently high testing frequency (a high penalty for mislabeling may not suffice).