Artificial intelligence (AI) in the service of the law holds out the promise of a legal system that is more accessible, effective, efficient and just, where artificially intelligent systems, widely accessible and equipped with perfect knowledge of the law and of all jurisprudence, assure high-quality legal representation and access to justice not just for the wealthiest among us but also for the most destitute and most vulnerable; where consistent outcomes, appropriate to specific circumstances, devoid of racial or other biases, are assured in courts across the nation; where the truth of facts in controversy in civil or criminal proceedings can be determined promptly and cheaply; and where resource allocation in law enforcement is self-optimizing to produce greater safety for all of us while concurrently protecting the civil rights of each of us.

But AI in the service of the law also brings the dystopian risk of a dehumanized legal system, where the functions of the law and the values that animate it are undermined by artificially intelligent caretakers, devoid of empathy, sacrificing the spirit of the law at the altar of efficiency, progressively corrupting principles, such as the presumption of innocence, fundamental to our legal and democratic order. Such risks may seem remote in societies with strong democratic traditions, where the rule of law is the most cherished of all institutions of society. But the adoption of AI in the US legal system, in particular in criminal justice, serves as a cautionary tale. In what may be characterized as an “efficiency trap,” AI is increasingly used to streamline bail hearing and sentencing processing in courtrooms across the United States, when the evidence supporting the effectiveness and fairness of these algorithms is, at best, highly contested.

How can societies seize the extraordinary potential benefits AI offers to legal systems while mitigating the risks its use in the creation, practice, and enforcement of the law poses to human dignity and well-being? How can policymakers, judges, lawyers, and other stakeholders in the legal system reliably decide what decisions to delegate to intelligent machines in the legal system? What institutional safeguards can be put in place to provide citizens with the grounds for trusting (or mistrusting) that the use of AI in the legal system in fact protects and advances the values that the law exists to maintain?