Police departments now record far more video than any team of people could ever watch. That footage becomes evidence for prosecutors, discovery for public defenders and, eventually, exhibits in court. Somebody has to review it, log it and share it on time, and most offices are short on staff.
That pressure explains why AI in criminal justice moved from pilot projects to signed contracts this year. Below are the developments that stood out in 2026, followed by how the technology is used across the system, where it fits into daily case work, and the benefits, risks and bias concerns that come with it.
What's New in AI for Criminal Justice
Body-Cam Footage Has Outgrown Human Review
Body cameras spread quickly after 2014. By 2020, about 79 percent of U.S. police officers worked in departments that used them, according to a research summary in The Regulatory Review. Axon, the largest supplier of police cameras and evidence storage, grew its stored footage from about 6 terabytes in 2016 to more than 100 petabytes, which ProPublica reported is enough for over 5,000 years of HD video.
Local numbers show the same gap. Police in Casper, Wyoming, collect about 60,000 hours of body-camera video a year and say they can review only around 1 percent of it, which is why the department has been testing AI review software, according to local news coverage. Tools in this category turn the audio into a transcript, label the type of encounter, such as a traffic stop, and flag moments a supervisor should look at, such as threats or use of force.
Prosecutors Are Buying AI for Discovery
In April 2026, Clackamas County, Oregon, took up a 10-year, $2 million contract with Axon that adds AI tools for the district attorney's office, OPB reported. A county report said the tools would support discovery, trial preparation and exhibits. Axon describes its AI as transcribing interviews and body-camera footage, surfacing key events and writing summaries.
Other prosecutors urged care. A deputy chief at the King County Prosecuting Attorney's Office told OPB that AI which points a person to a source they can check is less worrying than AI that recommends charges, which he said "raises more red flags."
Public Defenders Are Building Their Own Tools
In July 2026, the New Jersey Office of the Public Defender launched the OPD Resource Library, an AI platform built with researchers from Princeton University, the New Jersey Innovation Authority and the state's Office of Information Technology, the ABA Journal reported. It pulls years of briefs, motions, research and training material into one searchable place, so defenders spend less time hunting for documents and more time on strategy and clients. The office said commercial AI tools weren't designed for the language, reasoning or security needs of public defense, and New Jersey Public Defender Jen Sellitti said AI "should be making our lawyers more human, not less."
Statewide Prosecutor Offices Use AI for Research and Trial Prep
In February 2026, the director of the Tennessee District Attorneys General Conference told state lawmakers that DA offices across Tennessee use AI for administrative work, legal research and trial preparation, NewsChannel 9 reported. He linked it to staffing, since the state was about 75 assistant district attorneys short. A conference spokesperson was clear about the limit: AI "does not handle any cases in any capacity."
How Is AI Used in Criminal Justice?
A Stanford Law School summary of research prepared for the Council on Criminal Justice's task force on AI groups today's uses into three areas: police analysis of digital evidence and report drafting, prosecutorial discovery and charging support, and risk assessments and language models in the courts. In day-to-day work across the criminal justice system, that looks like this:
- Police: transcribing and flagging body-cam video, drafting reports from officer notes for the officer to correct and approve, and searching department policy.
- Prosecutors: summarizing police reports and footage, organizing discovery, and preparing exhibits and first drafts.
- Public defenders: working through discovery that can run to hundreds of hours of video and audio, and reusing past briefs and motions.
- Courts: risk assessment tools at pretrial and sentencing, and language models that judges and staff use to work through documents.
The common thread is volume. Most of these tools exist because the amount of digital evidence in a typical case has grown faster than the number of people available to review it.
Where AI Meets the Case File
An AI summary or a flagged clip only helps if it ends up in the right case, next to the charges, deadlines and evidence log. That is why much of the useful work in technology and criminal justice now happens inside case management systems rather than in separate AI apps.
Prosecutor offices, for example, run on prosecutor case management software that tracks each case from intake to disposition: police referrals, charging documents, discovery sent to the defense, witnesses and court dates. When AI output lands in that record, the reviewing attorney can check it against the source, and the office keeps one audit trail of what was produced and shared. When it sits in a separate tool, someone has to copy it over, and that is where errors and missed disclosures creep in.
Some case management vendors now build AI into the system itself. LegalEdge, for example, offers an assistant called LEA that summarizes police reports and case files and surfaces key moments in body-worn camera footage, hosted in AWS GovCloud inside the office's own environment. For agencies that handle criminal justice information, where the AI runs and who can see the data matter as much as what the AI can do.
Benefits of AI in Criminal Justice
- It covers footage nobody has time to watch. Reviewing 1 percent of body-cam video leaves 99 percent unseen. AI can at least point people to the moments that matter.
- It gives time back to short-staffed offices. Faster summaries and discovery prep matter most where offices are down dozens of attorneys, as Tennessee's prosecutors are.
- It can make work more consistent. The Stanford summary notes AI's potential to reduce backlogs and apply the same steps across large volumes of data.
- It keeps institutional knowledge in reach. New Jersey's library puts years of public defense work product one search away for every attorney in the office.
Disadvantages of AI in Criminal Justice
The same tools carry real risks, and the people closest to the work are the first to point them out.
- Errors can be hard to spot. Critics have said AI-written police reports contain mistakes both large and small, according to OPB, and a polished summary can hide a missing detail.
- Findings can stay hidden. ProPublica found that many departments keep AI review results confidential, and some, such as Seattle and Vallejo, California, cancelled contracts after union pushback.
- Buyers often lack the expertise to judge the tools. The Stanford summary warns that vendors market directly to agencies that may not have the technical staff to evaluate them.
- It can widen surveillance. The Regulatory Review piece raises concerns that camera data could be used to watch vulnerable and minority communities more closely.
AI Bias in Criminal Justice
Bias is the risk that draws the most attention. Tools trained on historical data can carry the patterns in that data forward, including uneven arrest and charging decisions. The King County deputy chief warned OPB that AI in criminal justice decisions could have a disproportionate impact on the people involved.
Bias can also come from what a tool measures. The Regulatory Review notes that body-cam analysis focused on surface language may miss a polite but unlawful stop, and that positive labels can nudge reviewers toward a favorable view of an encounter. So bias can show up in what a tool ignores as well as in who it flags.
What Good AI Governance Looks Like
None of the offices above are using AI to decide outcomes, and the safeguards they describe are a useful checklist for anyone buying these tools:
- A person reviews every output. Casper's officers must review, correct and approve AI-drafted reports, and Tennessee's DA offices keep every case decision with an attorney.
- Clear written rules for each use. Summaries, transcripts and search tools each need rules about when they can be relied on.
- Independent testing. The Stanford summary calls for a dedicated body to evaluate high-impact tools, publish benchmarks and track incidents, and The Regulatory Review urges vendors to publish what their labels measure.
- Training on limits and bias for everyone who uses the output, not only the technical staff.
- Secure hosting. Criminal justice information falls under the FBI's CJIS Security Policy, and agencies should know where their data is stored and confirm it is not used to train outside models.
Final Thoughts
AI in criminal justice is no longer a forecast. Police departments are using it to review footage, prosecutors are buying it for discovery, and public defenders are building their own tools. The offices getting the most from it treat AI as a fast first reader, keep a person responsible for every decision, and make sure its output lands in the case file where it can be checked.
The open questions are about rules rather than technology: who tests these tools, who sees the results, and how the public is told. The answers will decide whether this technology in criminal justice builds trust or erodes it.
Frequently Asked Questions
How is AI used in criminal justice?
Police use AI to transcribe and flag body-camera footage and draft reports. Prosecutors use it to summarize evidence and organize discovery. Public defenders use it to search large discovery sets and past work product, and courts use risk assessment tools and language models. In each case, a person is expected to review the output.
What are the benefits of AI in criminal justice?
The main benefits are speed and coverage. AI can review footage and documents that staff would never reach, cut the time spent on discovery and research, and make knowledge from past cases easier to find, which matters most in offices that are short on attorneys.
What are the disadvantages of AI in criminal justice?
The biggest concerns are errors that are hard to detect, bias carried over from historical data, results kept from the public, and wider surveillance. Agencies also often lack the technical staff to judge vendors' claims before they buy.
Does AI decide criminal cases?
No. The offices described here use AI for review, research and drafting, and keep charging, plea and case decisions with attorneys. Tennessee's district attorneys stated plainly that AI does not handle any cases.
