By Camello, M. L., Houston-Kolnik, J. D., & Planty, M

This technology brief explores the use of chatbots within the criminal justice system. The goal of this brief is to orient the reader to chatbots, present foundational insights from real-world examples of chatbot use, highlight considerations for implementation, and discuss the future of chatbots in the criminal justice system  

  Key Takeaways ¡ There are numerous benefits to implementing chatbots, including: Ÿ improved efficiency for users accessing information, Ÿ enhanced community engagement by creating a 24/7 communications channel, Ÿ expanded access to justice through multilingual chatbot capabilities, Ÿ reduced costs by automating FAQ support traditionally done through live chat, and Ÿ reduced staff workloads. ¡ Chatbots carry inherent risks that decision-makers need to consider before implementation, including: Ÿ misinterpretation of user input leading to incorrect responses, Ÿ biased training data, and Ÿ vulnerability to hacking. ¡ Advancements in AI have enhanced and will continue to enhance chatbot capabilities and applications; however, despite these advancements, deploying AI-driven chatbots is not a “plug-and-play” opportunity for criminal justice applications

Research Triangle Park, NC:RTI International, 2021. 15p.  

By Smith, J., Camello, M., & Planty, M

  Generative artificial intelligence (AI) refers to AI1 used to create content, such as text, images, music, audio, and videos.2 Generative AI offers many potential benefits, enabling users to automate, augment, and accelerate a wide range of workflows, from simple administrative tasks like transcription and translation to more-complex functions such as investigation and decision support. In the criminal justice system, generative AI offers promising solutions to address human resource and budget challenges, allowing practitioners to focus on more-impactful work. Generative AI–integrated tools may enhance data analysis, improve detection and objective assessment of evidence, and streamline administrative processes. However, its integration, particularly in the criminal justice domain, raises some concerns, including potential biases, privacy issues, and the need for rigorous oversight to ensure effective implementation. It is unclear whether these tools can deliver on their promised efficiencies in practice, as evidenced by early research evaluating time savings of implementing AI-assisted report writing software.3 These concerns highlight the necessity for addressing bias and accuracy, maintaining strict data privacy and security protocols, and promoting transparency and accountability in AI-driven decisions and processes. This report is intended to help criminal justice decision-makers do the following: ¡ Understand what generative AI is and how it relates to the criminal justice system ¡ Identify how generative AI may be applied to tasks and jobs within the criminal justice system and the potential benefits, realities, and limitations ¡ Consider the technical, operational, and governance factors that may influence adoption and implementation ¡ Understand what makes up the generative AI technology stack and how models can be trained Key Takeaways • Generative AI represents an acceleration and advancement in technological innovation that already impacts the criminal justice system and will continue to do so—it is no longer a question of if or when, but how and to what extent. • Generative AI–powered software tools may offer many potential benefits, such as improving efficiency and augmenting capabilities across an extremely broad set of applications for criminal justice system stakeholders. Although these products hold promise, little empirical evidence currently supports or refutes promised benefits from these products. • Generative AI models can be deployed in various forms, including cloud-based models that centralize data processing and federated models that enable decentralized training across multiple locations, preserving data privacy and enhancing security for sensitive criminal justice applications. • Decision-makers should be aware of the substantial technical, operational, and governance risks associated with generative AI– powered software tools prior to implementation. • Responsible use of generative AI requires addressing bias and accuracy concerns, maintaining strict data privacy and security protocols, adhering to ethics and legal standards, and promoting transparency and accountability in AI-driven decisions and processes. • Generative AI technology is evolving faster than the legal or policy environment for AI—the criminal justice community must be proactive and must implement robust internal training and policy frameworks rather than relying solely on external legal or regulatory guidance.  

Research Triangle Park, NC: RTI International, 2025 28p.

By Eleni Manis, PHD and Albert Fox Cahn, Esq

In this report, S.T.O.P. documents the New York City Police Department (NYPD)’s repeated failures to comply with New York City’s Public Oversight of Surveillance Technology Act (POST Act). Enacted in 2020, the POST Act is the first law to oversee the NYPD’s use of surveillance technology. A first attempt to regulate NYPD’s surveillance tools, the law only requires NYPD disclose its surveillance tools. As this report establishes, NYPD falls far short of the reporting norms set by other police departments subject to similar surveillance technology oversight laws. The report concludes by calling on the New York City Council to use its oversight authority to ensure that the bill is not ignored. S.T.O.P. also recommends that individual lawmakers and civil society organizations continue to evaluate potential litigation, seeking judicial intervention to compel the NYPD to comply with the POST Act.

By Mahima Arya, Juilee Shivalker, Maxwell Votey, Jackie Singh, Albert Fox Cahn, ESQ., Eleni Manis, PHD, MPA

Ultra-Wideband (“UWB”) is a short-range wireless technology somewhat like Bluetooth or Wi-Fi, but with superior locating abilities, enabling the highly accurate identification of an object’s position in three-dimensional space. UWB capabilities are now standard on many newer-model smartphones, allowing users to track UWB-enabled beacons from their smartphones and allowing vendors to leverage inbuilt UWB capabilities to create massive sensor networks using unwitting users’ mobile devices. This report focuses on privacy and anti-trust concerns surrounding UWB beacons, the tip of the iceberg of planned UWB applications. Because UWB’s locating abilities are so precise, beacons provide an easy way to track and stalk people. Beacons pass detailed device location data through neighboring devices’ networks, introducing the twin risks of malicious hacking and commercial exploitation by vendors. Apple and Amazon have acted to shut competitors out of the UWB beacon space, demonstrating a clear linkage between the story of these beacons and the larger story of Big Tech anti-trust concerns.

By Paula Garcia-Salazar, Nina Loshkajian, Albert Fox Cahn, ESQ., Eleni Manis, PHD, MPA

Everyone sees them: signs “welcoming” you to a neighborhood with the warning that “All Suspicious Persons and Activities Reported to Law Enforcement.” In the 1960’s, Neighborhood Watch groups proliferated in supposed response to increased burglaries. 1 In fact, the groups appear to have been a direct response to increased residential integration.2 A brainchild of the National Sheriffs Association, Neighborhood Watch groups were touted as a way to increase community involvement in crime prevention by encouraging residents to patrol their own streets and act as the eyes and ears of the local police. 3 But too often, local residents have interpreted this as a chance to become vigilantes, in many cases acting purely on bias to raise false alarms and profile fellow community members, endangering the very people these groups are allegedly designed to protect. 4 The groups have proliferated across the country even as they have been demonstrated to promote profiling and distract from actual public safety, there being little evidence that Neighborhood Watch programs reduce crime. 5 Now, a new form of Neighborhood Watch is here: smartphone-based apps that supplant the classic program with online forums for local neighborhoods. Nextdoor (27 million regular users),6 Neighbors by Ring (10 million users),7 Citizen (7 million users),8 and recently piloted Facebook Neighborhoods deliver “hyperlocal” updates to geographic “neighbors.” Nextdoor and Facebook Neighborhoods invite users to post on a range of local-interest content: upcoming events, business reviews, goods for sale, and so on. But the backbone of neighborhood surveillance apps is crime, both real and imagined. Apps encourage users to upload video footage, photos and descriptions of suspected crimes and supposedly “suspicious” people near their homes, producing reports riddled with dog whistles and overt racism. Apps court a police audience for these posts and enable police to request app users’ video footage, photos, and input. Apps even drive user engagement by inviting bystanders join in on the crime-oriented conversation: as on Facebook, users can comment, “like” and otherwise interact with posts with the click of a button.

By Albert Fox Cahn, ESQ. Caroline Magee

On January 7, 2020, New York City released its Internet Master Plan. The document identified how many New Yorkers lacked access to broadband and what the City intended to do about it. The numbers were staggering: 46% of New York households in poverty lack a home broadband subscription.[1]

But what had been a problem evolved into a crisis when the COVID-19 pandemic descended on New York City in March. For the first time, New Yorkers had to stay home: as New York’s 1.1 million public school students logged into Zoom for the first time, and their parents tried to take phone meetings in the same rooms, it became clear that the internet, once a luxury, was now a necessity.

As the fall semester loomed, the de Blasio administration tried to close the gap in July 2020, investing $157 million for providing low-or-no-cost internet to 600,000 New Yorkers, one-third of whom live in New York City Housing Authority housing.[2] The City is scrambling. In this light, a plan to expand internet access for residents of New York City is much needed and reflects the modern reality of reliable, affordable internet access as a barrier for reaching public services and economic opportunities. What is missing from the City’s Internet Master Plan, however, is a needed degree of specificity on the privacy and cybersecurity protections built into this planned internet expansion.

By S.T.O.P

In this first annual surveillance census, S.T.O.P. sought to map out all of the internet-enabled cameras operating in New York City. Even as many companies hide the location of their surveillance equipment, the Chinese-based firm Hikvision still allows their devices to be located…and the results are shocking. We identified 16,692 cameras in New Yorker City alone. This page details where Hikvision cameras across the five boroughs. While the numbers are extraordinarily high, please remember that for every one Hikvision camera we have mapped, there are dozens, possibly hundreds of other camera systems whose location remains hidden.

By S.T.O.P.

Our review of the 146 individuals who plead guilty in connection with the January 6th insurrection shows that facial recognition and other surveillance technology is not needed to properly identify suspects. Department of Justice data documents only 3 cases that used facial recognition. In contrast, the vast majority of cases used low-tech and less-invasive techniques, with 104 cases using tips from the public.

By Helen Webley-Brown, Anna Sipek, Katie Buoymaster, Juliee Shivalker, Will Owen, Eleni Manis, PHD, MPA

U.S. cities are squandering money on ShotSpotter’s unproven gunshot surveillance technology. 

• ShotSpotter surveillance increases police activity, but it wastes officers’ time. One major study of the technology showed that ShotSpotter fails as an investigative tool, providing no evidence of a gun-related crime more than 90% of the time and producing exceedingly few arrests (less than 1 per 200 stops) and recovered guns (less than 1 per 300 stops).

• ShotSpotter fails the Black and Latinx communities where it appears to be disproportionately deployed. The tool increases police activity and the risk of police violence without producing any significant effect on firearm offenses or on shooting victims’ medical outcomes.

By Evan Enzer, Anna Sipek, Mahima Arya, Nina Loshkajian, David Siffert, Eleni Manis, PHD, MPA

New cars are surveillance on wheels, sending sensitive passenger data to carmakers and police. Cars also store enormous amounts of passenger data onboard, where police can extract it using specialized tools. We estimate that law enforcement agencies could have accessed car data hundreds of thousands of times in 2020.

• Constitutional loopholes allow access to most data on cars without a warrant. Police can access information from car-connected phones and online accounts without the warrant typically required.

• U.S. immigration agencies weaponize car data. Other law enforcement agencies are poised to follow suit if they are not already doing so.

• New legislation, enforcement of existing data protection laws, and responsible car design and data storage policies can shift car data surveillance into reverse.

Police foundations allow police departments to secretly fund controversial programs and equipment.

• Foundations invest in dangerous surveillance tools like predictive policing software, digital surveillance platforms, cellphone hacking devices, and robotic spy dogs. 

• Foundations allow departments and officers to accept gifts from contractors in a way that would normally be illegal for city employees.

• Foundations violate good-government standards for city agencies and transparency standards for nonprofit organizations. Ideally, they should be abolished, but at a minimum, cities must end untraceable donations and corporate influence peddling.

By Evan Enzer, Arjun Ravi, Julian Melendi, Sohini Upadhyay, Eleni Manis, PHD, MPA

“Smart home” devices record audio and video in the home—and even collect daily schedules and health details. Once collected, this data is less than a warrant or data breach away from police and hacker access. Across the board, smart home devices have superior, privacy-protecting alternatives that perform the same key functions. The law doesn’t protect smart home users. “Do not buy” is the best advice until it does

Eleni Manis, PHD, MPA, Alexander Hughes, PHD

As NYC relies increasingly on traffic cameras, the NYPD has pulled over more and more drivers for a minor traffic infraction—license plate obstruction—particularly in precincts with the most BIPOC residents.

• This problem is only getting worse: NYPD’s racist enforcement gap doubled between 2016 and 2021.

• Automated traffic enforcement shouldn’t lead to more in-person traffic stops. Cities should study cameras' effects and adjust policing policies to ensure that cameras don't contribute to the over policing of BIPOC communities.

By Andy Ratto, Nino Loshkajian, Eleni Manis, PHD, MPA

• Police increasingly replace stop-and-frisk practices with databases that crudely profile Black and Latinx youth based on their neighborhoods, peer groups, and clothing.

• These databases ruin lives: police typecast minority youths as gang members without evidence, putting them at risk of false arrest and wrongful deportation.

• Many police departments refuse to implement due process safeguards despite clear evidence that their databases are based on racial profiling, not evidence.

• Even the most rigorous safeguards would be insufficient to mitigate the full range of harms that these databases pose. They must be eliminated in their entirety.

By Mina Kim, PhD; Phillip Atiba Solomon, PhD; Justin M. Feldman, ScD

Introduction In the US, injuries caused by law enforcement are a public health concern and driver of racial health inequities.1-4 While fatalities have attracted considerable public attention, nonfatal injuries inflicted by police are far more prevalent. Analyzing nonfatal injury trends can therefore help to answer critical questions about how populations experience policing, including whether the major advocacy efforts related to racial inequity in policing occurring from 2014 to 20215 coincided with changes to rates of legal intervention injury overall or by racial and ethnic group. Methods This repeated cross-sectional study analyzed publicly accessible, deidentified data from the National Electronic Injury Surveillance System—All Injuries Program (NEISS-AIP), a nationally representative sample of US hospital emergency departments (EDs), for the period 2004 to 2021. Local ethics review and informed consent were not required in accordance with the Common Rule. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies. In NEISS-AIP, legal intervention includes an injury or poisoning caused by on-duty police or other legal authorities, including private security guards. NEISS-AIP offers an advantage over administrative claims data, which underreport substantial shares of legal intervention injuries.6 We used NEISS-AIP’s predefined race and ethnicity categories, which were derived from patient medical records (eAppendix in Supplement 1). All analyses were performed between September 2024 and July 2025 using R software version 4.2 (R Project for Statistical Computing) with the survey and mgcv packages. We fit quasi-Poisson models for injury rates, treating year as a spline (for visualization) or as linear (to assess trends quantitatively), using parametric bootstrapping to construct CIs. We deemed any 95% CI for the linearized trend line that included the null value to be inconclusive as to its directionality, but we still interpreted the confidence limit as the bounds within which the trend was expect to fall. Additional methodological details are available in the eAppendix in Supplement 1.Results Between 2004 and 2021, a total of 1 500 577 ED visits (95% CI, 1 073 632-1 927 522; 85% [95% CI, 60% to 100%] men; mean [SD] age, 33 [12] years) in the US were for legal intervention injuries. Among patients with race and ethnicity data, 42.3% (95% CI, 21.2% to 63.4%) were African American or Black, 13.9% (95% CI, 5.4% to 22.3%) were Hispanic or Latinx, and 41.1% (95% CI, 30.7% to 51.5%) were White. Most patients were treated and released, with only 4.3% (95% CI, 2.5% to 6.1%) requiring hospitalization (Table). Legal intervention injury rates for the US population as a whole remained relatively stable over the study period (Figure), with the 2021 rate at 92% (95% CI, 71% to 119%) the level of the 2004 rate. Over the study period, mean injury rates for African American or Black people were 5.3 (95% CI,4.6 to 6.2) times those of White people. Rates for Hispanic or Latinx people were 1.5 (95% CI, 1.2 to 1.7) times those of White people. While the point estimates for Black:White RR decreased by 6%, from 5.48 in 2004 to 5.13 in 2021, uncertainty was high, with the 95% CI ranging from a 42% reduction to 51% increase. For the Latinx population, the RR decreased from 1.94 in 2004 to 1.06 in 2021, corresponding to a 45% (95% CI, 12% to 66%) decrease.

JAMA Netw Open. 2025

By Brandon Garrett

Most jurisdictions in the United States now record interrogations, including all federal law enforcement agencies, thirty states, and the District of Columbia. This memo provides updated information about the state-level adoption of electronic recording requirements, in a table displayed below, through statutes, court rulings and rules, and police policies. In addition, many localities now require the recording of some of all interrogations. 1 Leading law, psychology, criminal procedure, and policing experts have long recommended electronically recording interrogations, preferably by videotaping the full interrogation. The American Psychology-Law Society has set out what is known about police-induced confessions and has recommended mandatory electronic recording. 2 The International Association of Chiefs of Police recommends recording “all interviews involving major crimes” and prefers video recordings.3 The American Law Institute’s Principles of Policing set out the principle that: “Written policies should set out the procedures for the recording of questioning, and for the disclosure and the retention of recorded evidence.” 4 A large body of high-profile exonerations of innocent persons have occurred in cases in which false confessions were obtained during interrogations that were not recorded. In cases of “confession contamination,” law enforcement falsely reports that a suspect had volunteered supposedly inside information during an interrogation. Almost all of the persons exonerated by post-conviction DNA testing, who had falsely confessed, also had reports by law enforcement that the defendant had volunteered key crime scene details. We now know, with the benefit of DNA testing, that law enforcement contaminated these interrogations through the use of leading questions or feeding facts to the suspect. And none of the interrogations in those cases of DNA exonerations had been recorded in their entirety.5 Recording police questioning assists law-enforcement agencies by furthering the important goal of documenting evidence and ensuring the conviction of those who commit wrongdoing.6 Video recordings also empower judges to better assess the reliability of interrogation evidence, both to reject false claims of police overreaching and to examine potential wrongful convictions. 7 Agencies have reported positive experiences with recording interrogations because it provides powerful documentation that interrogations are conducted professionally and non-coercively.8 Fears that “few would allow themselves to be interviewed or interrogated” if it were known that interviews and interrogations are recorded have not been realized in jurisdictions in which recording has been introduced. 9 That said, the Principles of Policing noted that some flexibility with reluctant witnesses may be important. In addition, it may be increasingly feasible to conduct video recording in the field, as body-worn cameras are utilized more widely by agencies. And that said, states do often provide for exceptions due to exigent circumstances, equipment malfunction and the like.10 As summarized below, state statutes increasingly have required recording at least some categories of police questioning of Justice has a memorandum setting out a policy for recording interrogations; several states have done the same. Although the Principles of Policing took no position on the admissibility of unrecorded statements, others have. For example, the Alaska Supreme Court has ruled that judges should suppress unrecorded statements unless failure to record is excused by good cause; other courts have adopted court rules or rules of evidence requiring electronic recording of interrogations. 12 The table below summarizes these statutes, court rulings, rules, and policies.

Durham, NC: Wilson Center for Science and Justice Duke University School of Law. 2024. 7p.

By Marcus Pollard and Angie Weis Gammell

Guilt beyond a reasonable doubt is the bedrock principle of the American criminal legal system, and yet, thousands of innocent people have been convicted of crimes they did not commit. These wrongful convictions undermine the integrity of our criminal legal system. The National Registry of Exonerations (National Registry) has tracked every known exoneration since 1989 and reports that there have been 3,348 exonerations amounting to 29,950 years lost to wrongful incarceration. 1 This includes 75 exonerations in North Carolina since 1989, representing a combined 963 years of wrongful confinement. 2 In response to high-profile cases and exonerations, North Carolina implemented significant reforms in the early 2000s and became a leader in preventing wrongful convictions. During this time, North Carolina implemented open file sharing and established the North Carolina Innocence Inquiry Commission (“Commission”). North Carolina also enacted the Electronic Recording Act, the DNA Databank Act, and the Eyewitness Identification Reform Act. Several Innocence Project organizations were also established in North Carolina to represent individuals with wrongful convictions claims. One of those organizations, the Duke Wrongful Convictions Clinic, went on to become a founding member of the Innocence Network, which is a network of innocence organizations that work to combat wrongful convictions. 3 Since 2006, however, there have been 58 exonerations, 11 of which had their underlying conviction occurring after 2006, despite having these reforms in place. 4 These most recent wrongful convictions could result from the stall in reforms in North Carolina since the early 2000s.This report describes the wave of legislation that occurred in North Carolina in the early 2000s, considering both the improvements that were made and the gaps that remain. It then analyzes the 11 documented cases of wrongful convictions since the establishment of the Commission in 2006 to identify the reforms that North Carolina still needs. This report emphasizes that to reclaim its role as a national leader in conviction integrity, North Carolina should amend its open-file discovery policy, police interrogation practices, plea process, use of eyewitnesses and police informants, and use of forensic evidence. Finally, this report offers recommendations on ways to prevent wrongful convictions and make relief more accessible for those with plausible claims of factual innocence to ensure that those who have been wrongfully convicted have a genuine opportunity to prove their innocence. Overall, this report underscores the importance of taking proactive steps to prevent wrongful convictions and providing relief to those who have been wrongly convicted.

Durham, NC: Wilson Center for Science and Justice Duke University School of Law. 2024. 52p.

After 44 years in prison for a conviction of rape in the small town of Concord, North Carolina, Ronnie Long was released from prison in 2021. Four months later, the Governor granted him a full pardon on the grounds of innocence.1 That pardon made it possible to obtain compensation from a state commission, which awarded Long $750,000. This amount was the maximum available under state law that permitted compensation of $50,000 per year of incarceration, but with an upper limit of $750,000.2 After his exoneration, Long also filed a civil rights lawsuit in federal court. This federal case against the city settled for $22 million in 2024. Long also received $3 million from the North Carolina State Crime Lab as a result of its “role in hiding evidence from Mr. Long and his legal team that proved his innocence.”3 This substantial settlement made Long Compensating Exonerees in the United States highly unusual among exonerees in the United States. Most exonerees, if they are compensated, received something more along the lines of the $50,000 per year available under the state law that initially compensated Long, and nothing more. Many exonerees receive no compensation at all. Academics have long criticized the traditional lack of compensation for wrongful conviction in the United States.4 In 1932, Edwin Borchard wrote that the United States needed national legislation regarding compensation for wrongful convictions.5 For many decades, no such legislation existed, however. Civil rights litigation, like the case brought by Long, was uncommon, and compensation under statutory schemes was similarly rare. This changed once exonerations became much more common, including those based on post-conviction DNA evidence, in the 1990s. In general, many of the exonerees that have received substantial compensation, like Ronnie Long, have brought civil rights lawsuits in federal, not state court. For example, of the first 250 DNA exonerees, 60% received some type of compensation, and of those, half of them obtained it in federal court.6 However, exonerees are increasingly seeeking compensation under state compensation statutes. In this fact sheet, we describe how compensation for exonerees has evolved in the past several decades, including through successful litigation efforts and through the enactment of compensation legislation, in thirty-nine states, Washington D.C., and by the federal government.We summarize each of these state and federal statutes in the table at the end of this document. To date, 39 states have enacted such compensation statutes, in addition to Washington D.C. and the federal government. In addition, five of the remaining 11 states currently have pending legislation on the issue. The figure below shows which jurisdictions have these statutes currently, and which have these statutes currently, and which have legislation pending

Durham, NC: Wilson Center for Science and Justice Duke University School of Law, 2025. 20p.

By Jessica Gettleman and Tyler Kendall

The Wilson Center for Science and Justice at Duke Law has released Evaluation of Fayetteville’s ShotSpotter Installation: Results from the First 18 Months, an independent analysis of the gunshot detection system’s use in Fayetteville, North Carolina. The report was commissioned by the City of Fayetteville to examine how ShotSpotter operated across three designated coverage zones from September 2023 through March 2025.

The evaluation reviewed data from 911 firearm-related calls for service, ShotSpotter’s internal “Ground Truth Tracking Workbook,” and Fayetteville’s public Open Data Portal. The study compared patterns before and after installation, both inside and outside ShotSpotter zones.

Key findings include:

• Incidence trends: Gunshot-related incidents declined citywide beginning in 2022, consistent with national trends. Incident levels remained relatively stable within the ShotSpotter coverage zones during the evaluation period.

• Notification volume: While the 911 call volume related to gunshots decreased citywide over the period being evaluated, the Fayetteville Police Department received a significantly increased number of ShotSpotter gunshot-related notifications in the ShotSpotter coverage zones compared to an equivalent period of time prior to the evaluation period.

• Response times: Police were dispatched and arrived more quickly following ShotSpotter alerts compared to 911 calls. alone. However, the available data do not allow assessment of whether these faster responses affected investigative or victim outcomes. In addition, officers spent more time on scene when responding to incidents reported through both a ShotSpotter alert and 911 call compared to those reported through only one source.

• Investigation and victim outcomes: Evidence collection, victim identification, and arrests occured most frequently when ShotSpotter alerts were accompanied by 911 calls. ShotSpotter-only alerts produce comparatively fewer investigative or victim-related outcomes, reflecting challenges including possible false alarms.

• Resource use: Many ShotSpotter-only alerts involve detection of a small number of rounds or “probable gunfire” only. These are associated with lower productivity in terms of evidence collection and victim identification. Strategic prioritization of alerts—such as deprioritizing single-shot alerts lacking 911 confirmation—may improve efficient use of police resources.

The report also notes important limitations. The available data did not include comparable outcomes for 911-only calls, nor confirmation rates for whether reported gunshots were verified on scene. As a result, the evaluation cannot directly compare the relative accuracy or effectiveness of ShotSpotter alerts versus 911 calls.

Ultimately, ShotSpotter provides Fayetteville with increased numbers of alerts about possible gunfire incidents and facilitates faster police response times in targeted zones. However, its impact on reducing gun violence and improving investigation and victim outcomes is limited when alerts are unaccompanied by traditional 911 calls. The Wilson Center’s report does not make recommendations about whether Fayetteville should continue using ShotSpotter but provides data and analysis to inform that decision.

“Cities face difficult choices about public safety, and it is essential that those decisions are guided by data,” said Brandon Garrett, Faculty Director of the Wilson Center. “We are grateful to provide evidence-based research and evaluations that can help communities like Fayetteville weigh their options and determine which approaches work best for them.”

Durham, NC: Wilson Center for Science and Justice at Duke Law;, 2025 68p.

By Brandon Martinez, Rebecca Tublitz. Emily West

The majority of people in local jails around the country are awaiting their criminal trial. This means they have not been convicted of the crimes that brought them in. Many also do not pose a danger to public safety, nor a significant risk of flight, but remain in custody because they cannot afford bail or bond—making incarceration a fact of financial ability, not of safety.

In fact, keeping those in jail who could be better served in the community can cause long-term instability. Well-established research demonstrates the harms of pretrial detention, including worse criminal legal system outcomes, disruption in employment, poorer mental and physical health, and strained social and familial relationships. Unnecessarily overcrowding jails also means those who actually do need attention—especially for risk factors such as mental health, substance use, and housing instability—aren't given the resources they need to avoid landing back in jail.

Given this research, cities and counties involved in the John D. and Catherine T. MacArthur Foundation’s Safety and Justice Challenge (SJC) have engaged in collaborative, multi-agency efforts to safely shrink their local jail populations and increase equity across the system. This includes using resources to promote fairer and more just pretrial decision-making, address the needs of system-involved people, and establish a robust array of alternative options.

These efforts have yielded impressive results: in these SJC cities and counties, nearly 18,000 fewer people are in jails today compared to the start of the initiative. This reduction was driven largely by fewer people being booked into jail. Progress has been slower on reducing racial and ethnic disparities, though in many sites there are fewer people of color in jail than there were prior to the start of the initiative. Importantly, SJC research has established that throughout nearly a decade of the initiative, people released pretrial were no more likely to return to jail than before the SJC, including for violent crime—demonstrating that data-driven and cross-agency collaborative planning efforts can lead to positive outcomes for individuals while keeping communities safe.

Looking back at 10 years of work, CUNY ISLG analyzed individual-level jail data from five participating counties to explore trends in three key indicators:

1. Who gets booked into jail;

2. How they are released from custody (e.g., money bond, pretrial supervision);

3. How long they stay.

We also examine whether change in these domains varied for different racial and ethnic groups. The sites included in this analysis are Allegheny County, PA; Charleston County, SC; Palm Beach County, FL; New Orleans, LA; and Pima County, AZ. This brief presents the findings from the analysis.

New York: CUNY Institute for State and Local Governance. 2025.