Study Warns AI Tools in Criminal Justice Could Deepen Bias and Rights Violations

A new RAND Corporation and Council on Criminal Justice report warns that AI is rapidly transforming policing, courts, and corrections, often without proper oversight or safeguards. The study says predictive policing, surveillance, and risk assessment tools could reinforce racial bias and threaten civil liberties unless stronger transparency and accountability measures are introduced.

A major new report by the RAND Corporation and the Council on Criminal Justice warns that artificial intelligence is rapidly becoming part of everyday criminal justice operations in the United States, even as rules and safeguards struggle to keep pace. A study by the RAND Corporation examines how AI is now being used across policing, courts, corrections, and community supervision.

Researchers say AI is no longer experimental. It is already helping police predict crime hotspots, assisting courts with risk assessments, supporting prison monitoring systems, and automating administrative work. But the report argues that many agencies are adopting these tools without fully understanding their long-term impact on fairness, accountability, and civil liberties.

Predictive Tools Raise the Biggest Concerns

One of the report's strongest warnings focuses on predictive AI systems. These tools attempt to forecast outcomes such as recidivism, violence, or failure to appear in court. Systems like COMPAS and Public Safety Assessment are already used in many jurisdictions to guide decisions about bail, sentencing, and supervision.

Researchers warn that these systems often rely on historical criminal justice data, including arrest records and supervision histories, which may already contain racial and socioeconomic bias. As a result, AI can unintentionally repeat and even strengthen existing inequalities.

The report highlights past controversies involving predictive policing programs in cities such as Los Angeles and Chicago. In several cases, police departments faced criticism after AI systems repeatedly targeted already heavily policed neighborhoods, creating what researchers describe as "feedback loops" that reinforced biased enforcement patterns.

Surveillance Technology Is Expanding Fast

The study also shows how AI-powered surveillance is spreading quickly through law enforcement and corrections. Technologies now being used include facial recognition, biometric databases, automated license plate readers, geofencing tools, body-camera analytics, and AI systems that monitor phone calls and messages.

Researchers warn that these tools greatly increase the government's ability to track and monitor people. Facial recognition remains especially controversial because studies have shown higher error rates for Black individuals and women, contributing to documented cases of wrongful arrests.

The report also raises concerns about social media monitoring and AI systems capable of scanning inmate phone calls for suspicious language or behavior. Critics argue these technologies can blur the line between public safety and mass surveillance.

Courts and Lawyers Are Using AI Carefully

AI is also entering courtrooms and legal offices. Judges, prosecutors, and defense attorneys are beginning to use AI-powered systems for legal research, document review, evidence analysis, and case management.

However, the report notes that many judges remain cautious. Courts across the United States have started introducing rules requiring lawyers to verify AI-generated legal citations after several high-profile incidents where AI systems produced fake or inaccurate case references.

Researchers say the biggest risk may not be fully automated decision-making, but rather the subtle influence AI systems can have on human judgment. Even when judges or attorneys make final decisions themselves, algorithmic recommendations may shape how cases are viewed and handled.

RAND Calls for Stronger Oversight and Transparency

The report concludes that AI governance in criminal justice remains weak and inconsistent. Many agencies use systems that are difficult to audit or explain, while private technology vendors often refuse to reveal how their algorithms work.

To address these concerns, RAND recommends creating strict oversight rules for high-risk AI applications, especially those affecting liberty, sentencing, or surveillance. The report also calls for regular bias testing, stronger transparency standards, and more training so criminal justice professionals can understand the limits of AI systems.

Researchers stress that AI itself is not inherently good or bad. The real challenge, they argue, is ensuring that technologies designed to improve efficiency do not weaken fairness, accountability, or public trust. As AI becomes more deeply embedded in the justice system, the report warns that democratic oversight and human judgment must remain central to every major decision.