2025 Law Enforcement Misconduct Investigative Office Annual Report

Our previous annual reports have called for legislation to help localities establish specialized mental health units as alternatives to police response to situations involving mental health emergencies, and we repeat that recommendation here

Troubling statistics confirm the harms associated with the status quo. Studies indicate that people with mental illness generate one in ten calls for police service, and account for at least one in four fatal police encounters.[18] People with serious mental illness are more than eleven times as likely to experience police use of force, more than ten times as likely to be injured, and sixteen times more likely to be killed by law enforcement as those without serious mental illness.[19] Recent federal civil rights litigation has highlighted how failure to provide alternative, non-police responses to mental health emergencies could violate of anti-discrimination laws.[20] The 2025 Annual Report issued by OAG’s Office of Special Investigation, which investigates deaths caused by police or peace officers, highlights ten incidents that seemed to involve individuals in crisis who lost their lives during interactions with police.[21]

Many local governments, recognizing issues from relying on police as default first responders to mental health crises, are turning to co-responder models (where civilian specialists pair with police responders) or alternative first responder models (where armed police are not involved in the response).[22] As of September 2025, researchers at the Community Safety Workgroup have documented over 130 alternative crisis response units operating in the United States.[23] Cities developing new alternative crisis response models have achieved early successes. Eugene, Oregon, a pioneer of non-police response to mental health emergencies, adopted Crisis Assistance Helping Out on the Streets (CAHOOTS) in 1989, which led to responders requiring police backup in only 1% of mental health calls.[24] The city dispatches two-person teams of crisis workers and medics from Eugene’s White Bird Clinic to respond to 911 calls involving an individual in behavior health crisis.[25] As of 2019, the CAHOOTS team answered approximately 17% of the city’s police department calls and was estimated to save the city approximately $8.5 million in public safety spending annually.[26] In Durham, North Carolina, following the implementation of the Holistic Empathetic Assistance Response Team (HEART), a 2022 study documented a 34% drop in low-level crime in neighborhoods where the mental health responders operated in Denver, compared to neighborhoods that did not yet have responders.[27]

In 2021, the New Jersey Office of the Attorney General, in conjunction with New Jersey State Police and in consultation with the Department of Human Services, piloted an initiative called Alternative Responses to Reduce Instances of Violence and Escalation (ARRIVE) Together. The program paired State Troopers, who were trained in crisis intervention and de-escalation techniques, with a certified mental health screener and crisis specialist to respond to 911 calls involving behavioral health incidents.[28] A March 2023 report from the Brookings Institution evaluated data from ARRIVE’s first 14 months and found that the program reduced use of force, arrests, and racial disparities in the outcomes of behavioral health calls.[29] Of the 342 police service case calls evaluated in the report, 3% resulted in a use of force and seven incidents ended with an arrest. The use of social services also increased following the implementation of ARRIVE. Following a successful pilot, ARRIVE has been implemented in all 21 of New Jersey’s counties, with specific local implementation being tailored to the needs and resources of each community. Models include co-response, non-law enforcement response, law enforcement-assisted telehealth, and mental health professional follow-up subsequent to law enforcement response. From January 1 to November 3, 2025, ARRIVE teams responded to over 5,500 behavioral health calls throughout New Jersey, with only 1% of calls resulting in arrest and 33% resulting in a link to social services with the subject remaining in their community. 

In New York, the Daniel’s Law Task Force, established by state law in 2023 and charged with examining crisis response and identifying critical areas for improvement, published a detailed report in December 2024 that recommended (1) the development of a defined response protocol for behavioral health crises, including that health-led response teams should include mental health workers (including peers with lived experience), and that law enforcement should be deployed only where there is a threat of violence, and (2) the establishment of a Behavioral Crisis Technical Assistance Center that would develop standardized protocols, provide training and technical assistance to localities, and track and analyze data, and which would be guided by an advisory council of stakeholders that includes people with lived experience of the crisis response system.[30] So far, some localities in New York State already have started to develop alternative response programs, but assistance from the state will be crucial to expanding these efforts and providing funding and technical assistance.[31]

Accordingly, we strongly recommend legislation to help localities establish specialized mental health units that are specifically trained and equipped to provide assistance to those in mental health crisis, including by providing funding and technical support, consistent with the Daniel’s Law Task Force report. Such legislation should also require training for emergency dispatchers, create state and regional mental health councils to provide structure and oversight to local responses, and involve people with lived experience with behavioral health crises. For example, Daniel’s Law (named for Daniel Prude, who was experiencing a mental health crisis when he lost his life following an interaction with Rochester Police in March 2020), would create a statewide emergency and crisis response council to work in conjunction with the commissioners of mental health and addiction services to jointly approve emergency and crisis services plans submitted by local governments, as well as provide operational and financial support for emergency and crisis services provided to persons experiencing a mental health, alcohol use, or substance use crisis.[32]

The success of New Jersey’s ARRIVE Together program indicates that the implementation of statewide, community-tailored behavioral crisis response outlined in Daniel’s Law and the Daniel’s Law Task Force report is both possible and scalable for all of New York.

In addition, in 2025, New York enacted Executive Law § 840(8), which tasks the state Municipal Police Training Council (MPTC) and the Division of Criminal Justice Services (DCJS) with developing written policies for law enforcement agencies across New York State regarding response to situations involving individuals in behavioral health crisis who present a risk of harm to themselves or others, including techniques for de-escalation, minimizing use of force, and identifying alternatives to the criminal justice system. To ensure the successful implementation of these policies by covered agencies, the state should help agencies fund the improved level of training now required by Executive Law § 840(8), as recommended by OAG’s Office of Special Investigation.[33]

While the passage of Executive Law § 840(8) is a positive step, legislation that would help develop non-law enforcement responses to people with mental or behavioral health conditions has the potential to better allocate resources towards helping people in need of assistance and decrease the risk of harm from police encounters.

Law enforcement officers are empowered with broad authority to stop people, conduct searches, make arrests, and use force to do so. Yet despite these sensitive and sometimes dangerous activities, community members have limited access to data concerning interactions between law enforcement and the public. 

While New York has taken initial steps to improve law enforcement transparency, our laws do not cover many important indicators. With hundreds of law enforcement agencies in New York, to the extent that individual agencies separately collect and track data, the data is often inconsistent and hard to access.[34] As a result, the public and the state and local governments that represent them may be left in the dark when it comes to key information about policing and its impact.[35]

The data that is available, both in New York state and across the country, suggests that enforcement often disproportionately falls on people of color.[36]

There is a growing movement towards transparency, with nearly half of U.S. states now requiring data collection when an individual is stopped by law enforcement and most states prohibiting bias-based profiling.[37] New York should join these states by mandating uniform statewide data collection and by banning bias-based profiling by law enforcement.

One important component of such legislation would be to prohibit agencies from relying on race and other protected characteristics in initiating or determining the scope of law enforcement action, including pedestrian and traffic stops, searches, and arrests. This prohibition not only would bring New York in line with most other states, but also would permit New York to access a federal funding program to help support the collection and analysis of police traffic stop data as a means of reducing profiling.[38] To provide just a few examples, states like California,[39] Connecticut,[40] and Kansas[41] have all banned bias-based profiling in this manner.[42]

In addition, law enforcement agencies should track and report a standardized set of data on traffic and pedestrian stops including the duration and location of each encounter, the reason the encounter was initiated and its result, the perceived race, gender, and age of the person stopped, and actions taken by the officer during the encounter such as ordering a person out of a vehicle, handcuffing, searching people and vehicles, seizing property, making arrests, and using force.[43]

Statewide data transparency legislation should set a baseline, but should not impose limits on additional data that any localities wish to track and make public.

Law enforcement agencies should also make their policies publicly available online, except for policies that, if disclosed, would substantially undermine ongoing investigations or endanger officers or members of the public. They also should be required to publish collective bargaining agreements and disclose the amount they spend on settlements relating to alleged misconduct. 

Given the logistical challenges involved, data collection should be facilitated by a centralized state agency such as DCJS. DCJS would also periodically collect and publish key data reported by law enforcement agencies. These efforts could be supported by the federal funding that would be made available by the anti-profiling provision of the legislation.

Data collection and reporting systems have been implemented in states such as New Jersey[44] and California[45] and were recommended for police departments in the 2015 Final Report of the President’s Task Force on 21st Century Policing.[46]

In sum, legislation that bans profiling and requires agencies to track and publicize important data and policies would advance accountability and transparency, help agencies manage and evaluate their own performance, and further LEMIO’s statutory mission to enhance the effectiveness of law enforcement and protect civil liberties and civil rights.[47] 

Legislation to remove obstacles to strong independent civilian oversight could improve accountability and help prevent misconduct, particularly in the wake of the Court of Appeals’ recent ruling in Rochester Police Locust Club, Inc. v. City of Rochester[48] 

In 2020, Rochester residents voted to create a civilian Police Accountability Board (“PAB”) authorized, among other things, to discipline Rochester police officers for misconduct. In the Rochester Police Locust Club ruling, the Court of Appeals stripped the PAB of its disciplinary authority, holding that a 1985 change to municipal law in Rochester meant that, from that point forward, the discipline for Rochester police would be subject to collective bargaining with the local police union, regardless of any future municipal law-making or referenda. This ruling followed other Court of Appeals decisions in which the Court had concluded that the specific local legislative history of other municipalities (in particular, New York City, Schenectady, and Wallkill) meant that those other municipalities, unlike Rochester, retain control over discipline without union approval.[49] Several justices dissented from the Rochester Police Locust Club ruling, and both the majority and dissent noted that the ultimate resolution of these complex issues rests with the state legislature.[50]

Over the past several decades, many jurisdictions have established external civilian-led entities designed to provide independent oversight of law enforcement agencies.[51] The development of these civilian oversight entities generally has resulted from concerns over inadequate internal accountability processes and resulting breaches of public trust. Recently, prompted by increased scrutiny in the wake of high-profile incidents of law enforcement misconduct, civilian oversight entities have expanded—both in number and in the extent of their authority—in New York and across the country.[52]

As the ruling in Rochester Police Locust Club makes clear, despite the trend towards local civilian oversight, certain state laws have had the unintended consequence of restricting the development of civilian oversight entities. The civil service laws at issue in Rochester Police Locust Club are one example. As another example, Unconsolidated Law § 891 has been interpreted to prevent non-police personnel from presiding over police disciplinary hearings, unlike disciplinary hearings for other municipal employees.[53]

In the coming session, the legislature should address these issues and, more generally, facilitate civilian oversight by removing restrictions that prevent localities from determining the best way to establish police oversight and accountability mechanisms that serve the needs of their communities.[54]

Another way to remove obstacles to strong independent civilian oversight is to improve civilian oversight agencies’ access to sealed records. New Yorkers are protected by laws that restrict access to—or “seal”—certain criminal justice records maintained by law enforcement agencies and the courts. These sealing laws serve important functions, but they can unintentionally hinder investigations of law enforcement misconduct. LEMIO and oversight entities such as civilian review boards often require access to sealed criminal justice records to fully investigate allegations of misconduct. For example, investigators may need to review body-worn camera footage, arrest records, and other documents related to incidents involving police interaction with young people or cases in which charges were dismissed—yet access to these records may be blocked by sealing laws (although Executive Law § 75 permits the OAG to access sealed records maintained by covered agencies). 

Legislation to address this problem should amend various provisions of the Criminal Procedure Law, Civil Rights Law, and Family Court Act to make clear that the OAG, as well as other agencies investigating law enforcement misconduct, can access sealed records as needed.

Traffic stops may seem commonplace, but they can have serious consequences that range from the financial burdens of tickets to lengthy roadside detentions and searches. Law enforcement officers make more than a million traffic stops each year in New York.[55] These stops may involve speeding and safety-related issues, but many stops are based on Vehicle and Traffic Law (“VTL”) violations that are less critical to safety, such as an expired registration or equipment violations like a broken taillight or tinted windows.

Under current law, these traffic stops can be “pretextual” in nature. In other words, the primary reason for the stop may not be the potential VTL violation, but rather the officer’s hunch—which on its own would not meet the constitutional standard—that stopping the car might lead to evidence of another offense.[56] Given the wide range of potential VTL violations, it is not hard to find a reason to stop a car and hold its owner on the side of the road.

The racial disparities are significant. Black drivers are far more likely to be pulled over than white drivers.[57] While most New York agencies do not track searches and their outcomes, analyses from other states that do track data show that while drivers of color are more likely to be searched, the likelihood of officers finding contraband from those searches is much lower than for white drivers who are searched.[58] And, tragically, some of the high-profile deaths of individuals at the hands of police over the past decade originated from simple traffic stops that escalated into violence or high-speed pursuits.[59]

Some states and localities have attempted to mitigate the disparities and harmful outcomes from traffic stops by limiting stops for low-level equipment violations, repealing excess fees and implementing voucher programs to help fix equipment issues, and restricting high-speed pursuits to stops involving specific serious offenses. New York should consider legislation that adopts these reforms across the state.

Jurisdictions that have banned police stops for low-level equipment violations include states such as Virginia and Oregon, localities such as Philadelphia, Pittsburgh, and Memphis, and various local police departments across the country.[60] These jurisdictions have met successes: in Philadelphia, Pennsylvania, evidence suggested that officers were able to focus on serious public safety issues and get more guns off the street when they could divert time and resources away from low-level violations that fail to make the roadways safer.[61]Similarly, in Fayetteville, North Carolina, after police stopped conducting stops for certain low-level equipment violations, car accidents decreased and crime was unaffected, while racial disparities in traffic stops were reduced.[62] Components of such legislation include prohibiting police officers from initiating traffic stops for specified low-level equipment and traffic violations, rendering inadmissible in criminal actions any evidence derived from a traffic stop initiated on prohibited grounds, and requiring robust data collection on traffic stop encounters.

Jurisdictions across New York and the country have also begun to address the financial burdens of fines and fees associated with traffic stops.[63] As a report by the Vera Institute found, even low-level violations come with high financial costs: a single window tint violation can cost $340 in fines, fees, and surcharges in New York.[64] The cost of multiple traffic tickets can range in the thousands, a cost that many households cannot afford.[65] Some localities in New York, such as Buffalo, have repealed local fees associated with traffic infractions in light of the financial hardship they impose on residents.[66] Several localities across the state have also implemented voucher programs, where community groups, government agencies, or police officers issue vouchers that cover the costs of fixing equipment violations such as broken headlights and taillights, or have implemented programs waiving fines if the driver can show the equipment is fixed.[67] New York should consider legislation that eliminates the use of fees and surcharges imposed on those convicted of offenses under the penal and vehicle and traffic laws,[68] and should consider implementing the issuance of vouchers by community groups and government agencies for low-level equipment violations statewide.

Finally, New York State should join the growing number of municipalities and law enforcement agencies across the country that have begun to address the dangers of high-speed vehicle pursuits. A recent national study found that police pursuits resulted in the death of more than 3,000 people from 2017 to 2022, more than 500 of whom were bystanders, and that the vast majority of police vehicle pursuits were to stop drivers suspected of committing minor vehicle and traffic violations or lower-level crimes.[69] In Buffalo, pursuits are banned where the reason for the attempted stop is a Vehicle and Traffic Law violation, a misdemeanor, or a non-violent felony.[70] In Albany, officers are required to consider several factors before starting a chase, such as the seriousness of the offense for which the person is pursued, the present danger the fleeing person poses to the public, and the possibility of apprehending the person at a later time.[71] As detailed in a recent OAG report, New York law enforcement agencies should be required to track and publish a standardized set of data about traffic pursuits and high speed chases, facilitated by a centralized agency such as the Division of Criminal Justice Services (DCJS).[72] Statewide legislation should also ban high-speed pursuits, with narrow exceptions for situations when a serious or violent felony has been committed, and when the driver’s conduct poses an imminent threat of death.

The capabilities and use of artificial intelligence and machine learning technologies have increased dramatically in recent years. While these technologies are relatively new, law enforcement agencies have started to integrate them into their work. For example, these tools have been used for facial recognition, vehicle surveillance in the form of automatic license plate readers (ALPRs), and audio-based gunshot detection systems.[73] More recently, generative AI tools have been developed to produce audio/video transcriptions and draft police reports.[74]

While AI technologies may have important benefits, the unregulated use of AI and technology-assisted surveillance by law enforcement carries significant risks. For example, advocates have raised privacy concerns with the use of ALPRs to track and analyze the movements of millions of people.[75] The accuracy of facial recognition technology has been found to vary across developers and appears to exhibit gender and skin color biases that disproportionately place women, children, and people of color at risk of false positives.[76] In February 2024, the Innocence Project identified at least seven confirmed cases of misidentification due to the use of facial recognition, six of which involved Black people being wrongfully accused.[77] In January 2024 in Chicago, police opened fire on an underage boy after ShotSpotter, a gunshot detection acoustic surveillance tool, wrongfully identified the boy’s fireworks use as the sound of a gunshot.[78] In October 2025, an AI-powered weapons detection system at a Baltimore high school mistakenly identified a student’s bag of potato chips as a gun, resulting in police responding to the school and handcuffing the student.[79] The generative AI tools used to produce police reports rely on large language models that are prone to “hallucinations” — unpredictable errors in which the model misrepresents or fabricates facts. Many of these technologies lack mechanisms for auditing their inner workings to understand and evaluate their accuracy and performance.[80] Last month, a federal district judge raised concerns regarding the use of generative AI by federal law enforcement in use of force reports, observing that such use may have led to inaccuracies.[81] Based on similar concerns, in 2024, the King County Prosecuting Attorney’s Office in Washington circulated a memo informing law enforcement agencies that it would not accept AI-generated reports due to the software’s general lack of compliance with federal regulations and unaddressed concerns about the efficacy of the products.[82]

Some states have already passed legislation regulating various AI and technology-assisted surveillance tools, including regulation of facial recognition in Montana and Maryland, and ALPR regulation in Virginia and Minnesota.[83] Other states have required general standards for use of AI systems in law enforcement applications.[84] The New York City Council, in 2020, enacted the Public Oversight of Surveillance Technology (POST) Act, which requires the reporting and evaluation of surveillance technologies used by the NYPD, including facial recognition, license plate readers, and drone usage.[85]

As a first step towards ensuring that AI is responsibly utilized in law enforcement applications at the state level, our office recommends legislation that would require police agencies to conduct an inventory of, and develop a publicly available policy for, any AI used to aid in criminal investigations. Legislation should also require law enforcement agencies to disclose the use of AI in their police reports, which would ensure that prosecutors have access to information about police AI use in order to comply with disclosure obligations to criminal defendants and promote public trust. The Policing Project has drafted a model statute regarding police disclosure of AI to further these goals.[86]

The increased use of AI across all sectors, including law enforcement, appears inevitable. Ultimately, as the use of AI tools expands, more stringent regulation will be needed to ensure the accuracy of these technologies, account for bias, and allow for auditing of their inner workings. In the interim, basic transparency regarding the use of AI and surveillance technology by law enforcement is a crucial first step towards reducing the potential for harm and ensuring they are used responsibly.

Video footage of interactions between law enforcement and the public can promote accountability for misconduct that might otherwise go undetected and protect officers who are wrongly accused. It can also be a helpful management tool for improving performance and training. 

The legislature should require by statute that all police and sheriff’s departments deploy and use body-worn cameras and dashboard cameras in all encounters with members of the public, as recommended LEMIO’s prior annual reports and in the October 1, 2025 Annual Report of OAG’s Office of Special Investigation.[87] The statute should provide guidelines for retention of camera footage and permit departments to create exceptions to deployment or use, but any such exceptions must be limited to specific and clearly articulated and justified law enforcement need. Importantly, the state should provide smaller departments with funding as needed for such deployment and training.