Persons with cognitive impairment(s) are overrepresented in the criminal justice system (CJS) yet many instances of cognitive impairment go undiagnosed. As this article outlines, it would be both desirable and feasible to use automated alerts to flag accused persons who may require assistance in interacting with the CJS either due to a confirmed or likely diagnosis of a cognitive impairment or other relevant condition. A proposed method to develop this alert system is outlined, combining Natural Language Processing (NLP) applied to Electronic Health Records (EHRs) with data linkage (DL) of health and CJS data. Although there are technical barriers, this article focuses on the ethical and legal barriers of this proposed approach. It is concluded that the overall benefits of the proposed alert system would likely outweigh potential adverse outcomes. It is argued that a waiver of consent would be appropriate and that legal barriers, in terms of privacy legislation at both Federal and State levels, which apply varying requirements for the disclosure of personal information, may be overcome in part through de-identification strategies. The examples provided in this paper of criminological data linkage projects support the feasibility of the method proposed.