Resolution Supporting Responsible Evaluation and Use of Artificial Intelligence in Criminal Justice

Model Policy

WHEREAS, artificial intelligence (AI) offers opportunities for communities to improve public safety, strengthen the administration of justice, and support the efficient use of taxpayer resources; and

WHEREAS, AI use in criminal justice can affect individual liberty, due process, privacy, and public confidence, requiring careful consideration of the benefits, limitations, and potential harms of such use; and

WHEREAS, decisions about whether, when, and how to use AI tools in criminal justice should reflect the specific purpose, evidence, risks, and circumstances of each proposed application; and

WHEREAS, AI tools can now create, generate, alter, or enhance photographs, video, audio, and other material in ways that may be difficult or impossible for a judge or jury to detect, and the authenticity and reliability of evidence bear directly on the accuracy of verdicts and the fairness of criminal proceedings; and

WHEREAS, compliance with applicable law is essential but does not replace the responsibility of jurisdictions and agencies to evaluate the suitability of AI tools in criminal justice and maintain accountability for their use;

NOW, THEREFORE, BE IT RESOLVED, that the {Insert legislative body} encourages state and local criminal justice jurisdictions and agencies to undertake a structured, rigorous, and evidence-based assessment of AI tools before adopting or materially expanding their use in criminal justice, and to adopt safeguards proportionate to the risks of harm and potential consequences; and

BE IT FURTHER RESOLVED, that such assessments should identify the problem the AI tool is meant to address; compare such AI tool with existing practices and other alternatives; examine the evidence of the tool’s accuracy, reliability, effectiveness, costs, and impacts across the affected population and specific demographic groups; and determine whether the deploying agency has the capacity to implement and oversee the proposed use; and

BE IT FURTHER RESOLVED, that jurisdictions and agencies should establish clear responsibility for and meaningful human oversight of the AI tools adopted, and as necessary, protect sensitive information, train personnel, seek input from affected communities and justice partners, and secure contractual provisions supporting evaluation, accountability, and termination; and

BE IT FURTHER RESOLVED, that jurisdictions and agencies should monitor outcomes and reassess AI tools throughout their use, with clear criteria for piloting, expanding, modifying, suspending, or discontinuing an application, and recognize that declining or delaying adoption may be appropriate when evidence or safeguards are insufficient; and

BE IT FURTHER RESOLVED, that the {Insert legislative body} supports requiring that, before any evidence is admitted in a criminal proceeding, the party offering the evidence disclose to the court and to all other parties, within a reasonable time to permit meaningful review, whether AI was used to create, generate, alter, or enhance that evidence, including any photograph, video, or audio recording, and if so, identify the AI tool used and describe the nature and extent of its use, so that the court may determine the authenticity and reliability of the evidence before it is presented to the finder of fact; provided that such disclosure need not extend to routine, automated processing performed by a recording device at the time of capture that does not materially alter the content of the evidence; and

BE IT FURTHER RESOLVED, that jurisdictions and agencies are encouraged to consider practical resources, including the Council on Criminal Justice Task Force on Artificial Intelligence’s User Decision Framework, to support these assessments and ongoing oversight.