We present a model in which firms differing in creativity decide whether to invest in genuine research or to submit "bogus" patent applications (claiming that they invented something which is not truly novel). The government has to delegate the verification of novelty to an agency which must exert costly effort in order to obtain a signal of patentability. Firms self-select according to their creativity, with high-creativity types producing true innovations and low-creativity types submitting bogus applications, or staying idle. The thresholds depend on the expected examination effort and on the application fee. We show that, at the full-commitment optimum, all bogus applications are deterred. When the agency lacks commitment power and its effort is unobservable, the outcome hinges on whether the patentability signal is hard or soft information. With hard information, the government rewards the agency for rejecting patent applications and can attain an allocation that is arbitrarily close to the optimum, consistent with results from the literature on optimal audits. With soft information, however, transfers are constrained by the need to insure truthfulness, creating a tradeoff between innovation and deterrence.