Artificial intelligence (AI) makes it possible for machines to learn. AI can be used for good (e.g., more accurate IOD diagnoses by police physicians; data analysis of crime hot spots). It can also turn malevolent in both fiction (e.g., HAL 9000 in Stanley Kubrick’s 2001: A Space Odyssey) and in real life (e.g., self-driving car runs over pedestrian in San Francisco) because machines (like guns) are emotionless and without inherent ethics. What people choose to do with it matters, and it should be respected for its potential harm. Current applications of AI in law enforcement and policing are highlighted, and future considerations are discussed.

Automated license plate readers

Automated license plate readers, ALPRs or simply LPRs, allow for more effective proactive and reactive policing through AI technology. AI captures the image of the vehicle and its license plate, transforms the plate image into alphanumeric characters using optical character recognition, compares the plate number acquired to one or more databases (aka “hot lists”) and then automatically alerts law enforcement. Consider the following key areas where LPRs have helped law enforcement.

• Community caretaking. LPRs have been used to locate missing and endangered persons. Examples have included kidnapping cases (e.g., Livermore, California), mentally ill persons in crisis (e.g., Mount Juliet P.D. in Tennessee) and deceased persons (e.g., Grand Valley State University P.D. in Michigan).
• Investigation. LPRs have been used in policing to apprehend known suspects (e.g., murder suspect caught by Beverly Hills P.D.), identify previously unknown suspects (e.g., smash-and-grab arrests caught by San Francisco P.D.; homicide case solved by California Highway Patrol), recover stolen vehicles (e.g., New Orleans P.D.) and collect evidence (e.g., Katy P.D. in Texas; Gainesville P.D. in Florida).
• Crime prevention. LPRs have been used in proactive policing (e.g., University of Memphis P.D. in Tennessee) and personal protection (e.g., Bar Harbor P.D. in Maine). Sometimes just knowing that LPRs are in a community prevents offenders from committing crimes in that area.
• Traffic compliance. LPRs have been used to enforce parking and tolls (e.g., Metropolitan Transportation Authority in New York), enforce citations and fines, such as speeding or driving through a red light, speeding in a school zone (e.g., Bibb County Sheriff’s Office in Georgia) or just analyzing traffic-related activity (e.g., Belle Meade P.D. in Tennessee).

Facial recognition

Facial recognition is like finding a needle in a haystack. The software can find one face in a billion in about one-third of a second. It works by finding points on your face (e.g., eyes, nose, mouth) and measuring the distance between those points, and comparing that data to other faces either in video or still images.

Consider the case of a missing California woman who had gone to meet a man (whom she had met online) and never came home. When the complainant (i.e., the missing woman’s daughter) called to check on her, the male suspect made suspicious comments, and he would not allow the complainant to speak with the missing woman. The missing woman could be heard crying in the background. The complainant located the website, and although there was no information on the male’s identity on the website, there were a few good quality photos. Officers ran one of the photos through the facial recognition system. This led to identifying the suspect, including his license plate. The license plate was then entered into the LPR system, and the vehicle was quickly located. As a result, police located the woman (who had been drugged) and safely rescued her.

Advertisement analysis

AI machine-learning algorithms have been used to automatically analyze online advertisements (e.g., commercial sex ads) to determine if they are associated with human trafficking and if they belong to the same organization. AI technology will analyze semantic similarities in the ad descriptions, extracting unique phrases in content, and then finding matches across hundreds of thousands of ads across websites. Human trafficking networks are often elaborate, large scale and connected to each other, and AI allows law enforcement to disrupt the entirety of the network (as opposed to pieces).

Shots fired and weapons detection

Gunshot detection technology is being used in law enforcement as part of a comprehensive gun crime prevention strategy. Sophisticated machine algorithms identify gunfire and filter out firecrackers, car backfires and other urban noises. Although the technology, of course, does not prevent gun violence, and there is no silver bullet to address the problem of gun violence, the technology continues to expand in certain cities (e.g., Cleveland, Detroit).

Acoustic sensors have been placed in neighborhoods (e.g., Seattle, Portland, Detroit) with certain software to identify loud noises that are then automatically transmitted for analysis. If the noise is identified as gunfire, techs alert the local police department, and officers are dispatched immediately (even before 9-1-1 is dialed). This speeds police response. It also catches incidents where community residents may not have dialed 9-1-1 out of fear. As a result of the technology, some cities have cited a drop in homicides (e.g., Miami) and a reduction in violent crimes (e.g., Las Vegas). Chicago P.D. consistently describes gunshot detection technology as an important part of their operations.

Redaction capabilities

AI has been used in law enforcement agencies to remove certain identifying information from electronic police reports (e.g., suspect’s name, race, hair and eye color; specific neighborhoods or districts that might reflect race of those involved). This AI technology was introduced with the intent of minimizing disparate outcomes due to possible racial bias. Prosecutors then make an initial charging decision based on the AI-redacted police report.

AI technology can also redact or blur identifying information (e.g., faces) or sensitive footage (e.g., gross or demeaning content, or information that would do irreparable harm to innocent persons or their character if released). This lowers the workload for prepping for court cases and responding to public information requests. For example, it might only take “one touch” on AI redaction to handle a 30-second body camera clip instead of hours of editing.

Automated speech recognition

Voice recognition outperforms humans in court when identifying unfamiliar speakers. AI technology analyzes verbal statements and the idiosyncrasies of the speaker’s cadence, tone and other identifying markers to accurately indicate whether the speaker on an audio recording is the defendant.

CAD, predictive policing and chatbots

Law enforcement agencies have been using AI-enhanced CAD systems to better capture data, make more effective decisions on resource staffing and deployment, and automate workflow. AI-supported predictive policing models help identify those most at risk of being involved in a crime. AI-powered chatbots have been fielding routine 3-1-1 questions, providing emergency alerts and other important notifications to residents.

Future considerations

AI continues to expand law enforcement capabilities and efficiency. But, in turn, it also continues to expand criminal capabilities and efficiency.

AI-assisted autonomous weapon systems in the U.S. military do not require our military men and women to be placed in harm’s way. Might this be the future of American policing? Consider, for example, how Dallas SWAT killed the sniper who ambushed and murdered five police officers on July 7, 2016. That was the first time in U.S. history that a robot was used to kill a suspect, and it won’t be the last.

AI-assisted technologies in the military are also improving mission efficacy. For example, AI can be leveraged to quickly sift and sort copious amounts of data to give analysts time to apply critical thinking skills to achieve better mission outcomes, and the race is on to secure and reinforce such advantages over opponents. Consider China (AI), Russia (hypersonics) and Turkey (drone technology). The complexity of data and levels is not just additive; it’s multiplicative and dimensional. Might this be the future of American policing? Remember, AI is emotionless and without inherent ethics; it is the officer who thinks in a contextual and nuanced way. It is the officer who is entrusted to safeguard the humanity in each community. It is the officer who gives citizens hope for tomorrow.