Artificial Influence: AI and Discrimination in America’s Criminal Justice System
Author: Becca Stachel
September 10, 2026
![[Image credit: Ahmad Juliyanto/Vecteezy]](https://static.wixstatic.com/media/f05ed1_c575fac234844417bfb2e752a3cecbc8~mv2.jpg/v1/fill/w_980,h_549,al_c,q_85,usm_0.66_1.00_0.01,enc_avif,quality_auto/f05ed1_c575fac234844417bfb2e752a3cecbc8~mv2.jpg)
Interest in how the American criminal justice system treats Americans of different circumstances is not at all new to the modern age of artificial intelligence (AI) utilization–if anything, the very opposite is true. The American criminal justice system has been studied since its inception, due in some part to its vast, complex, and fragmented amalgamation of colonial British law and the development of new forms of legal thought over the past two hundred years. Over time, the system has morphed into the massive network of law enforcement, judicial proceedings, and correctional facilities, both public and private, that Americans encounter today. This system came about via the establishment of a set of legal proceedings spun off from the British system, which the forefathers of the United States (US) used as a foundation to create their own criminal justice framework. This is in part attributed to the dissatisfaction of the settlers with certain rules that had prompted their exodus in the first place. Within its colonial status, American policing as a public service did not exist, nor did the appeals system, and courts retained no specialization nor differentiation, unlike the variety of specific courts that operate today; they processed all cases, from juvenile to narcotics-related crime.¹ As the US established its own legal code and enforcement structure as part of independence, prisons were established and frequently held prisoners in appalling conditions, which, although not unique to America, prompted the nation’s first calls for prison reform.² The 1950s saw a decrease in the national murder rate, which gave rise to political support for reform and rehabilitation as opposed to punitive incarceration. In 1954, the American Prison Association changed its name to the American Correctional Association, a major change in language indicative of how tides were beginning to shift regarding how Americans viewed the impetus of the criminal justice system.³ Throughout the Civil Rights movement, theories of justice developed in tandem with the structural work being done to address racial discrimination and inequitable policing tactics; these theories focused on a language supporting the practice of upholding human rights, the creation of operational standards for law enforcement, and a push towards the study of incarceration as a system. Such shifts in thinking about justice reflected the first development in theorizing the purpose of the American criminal justice system. As seen in such philosophies as the ‘justice as fairness’ theory of political philosopher John Rawls, justice in Western political thought is built upon the notion of a society of free citizens holding equal rights to one another and cooperating within an egalitarian economy.⁴ This essay explores both rehabilitation and restitutive justice in the American criminal justice system whilst explaining how 21st-century technology, including the development and widespread usage of artificial intelligence (AI) technology, has disproportionately impacted the interactions of individuals of color with policing, incarceration, and rehabilitation and re-entry into a discriminatory and tech-forward society.
Despite initial steps towards prison reform, the march towards progressive change in the criminal justice system staggered after New York sociologist Robert Martinson penned a damning report on prison rehabilitation.⁵ The report, harsh in tone and pessimistic with regard to the future potential of many incarcerated individuals, was followed by the 1989 Supreme Court case Mistretta v. United States. In Mistretta v. United States, the Supreme Court upheld federal ‘sentencing guidelines’ that removed the consideration of rehabilitation as a sentence for convicted individuals.⁶ Despite this ruling and a series of large blows dealt to the prison reform movement in the second half of the century, those within the movement continued to push for reform, if not rehabilitation. Today, prison reform is still a pressing talking point in the nation’s domestic political discourse, and individuals on both sides of the debate have remained animated in their fight over this aspect of the carceral system. Tensions run high in debates on restitutive justice initiatives and the impact of stigma surrounding incarcerated individuals, especially when taking into consideration both the impact of systemic racism in the US and the role that modern technological advancements play in exacerbating the racial inequity and discrimination already prevalent in American society.
To discuss the implications of artificial intelligence on the American justice system, it is imperative to understand the pervasiveness of systemic racism within it. In 2023, the United Nations (UN) Office of the High Commissioner of Human Rights published a press release titled “Systemic racism pervades US police and justice systems, UN Mechanism on Racial Justice in Law Enforcement says in new report urging reform.”⁷ The report mentioned in the press release was penned by the UN International Independent Expert Mechanism to Advance Racial Justice and Equality in the Context of Law Enforcement (Expert Mechanism) following a visit to the US.⁸ UN experts supported the notion that systemic racism is a pervasive issue in American police forces and the criminal justice system, and the Expert Mechanism subsequently urged the US to address these problems and put effort towards reform. During the Expert Mechanism’s visit to the US, it heard testimonies from 133 affected individuals, visited five detention centers, and held meetings with a range of governmental, law enforcement, and civilian groups in six major cities. The report found that “racism in the US - a legacy of slavery, the slave trade, and one hundred years of legalized apartheid that followed slavery’s abolition - continues to exist today in the form of racial profiling, police killings, and many other human rights violations.”⁹ Statistically speaking, more than one thousand individuals are killed by American police forces each year, with only 1% of these killings resulting in officers being charged.¹⁰ The UN recommended that use of force regulations be reformed to abide by international standards to decrease the number of killings. Additionally, the report cited “with profound concern” instances where Black children are sentenced to life imprisonment, incarcerated pregnant women being chained during childbirth, and inmates being subjected to inhumane durations of solitary confinement—the Expert Mechanism is clear about the “appalling overrepresentation of people of African descent in the criminal justice system.” This press release and the adjoining report underscore the importance of accounting for racism when discussing the criminal justice system in the US. Additionally, in another recent report from the same UN entity, the Special Rapporteur of the Human Rights Council states, “Recent developments in generative artificial intelligence and the burgeoning application of artificial intelligence continue to raise serious human rights issues, including concerns about racial discrimination.”¹¹ Overpolicing in Black communities, however, yields a negative feedback loop, according to the findings in the article. This is only further reinforced by AI algorithms which, when officers record new offenses, are fed information that generates biased predictions that target the exact neighborhoods already victim to overpolicing. As these neighborhoods are targeted, predictions continue to increase. Systemic racism thus taints the criminal justice system by skewing the predictive ability of the algorithms.
It is evident that racism is a deep-seated, systemic issue in the American criminal justice system and in policing. Given that artificial intelligence raises concerns about racial discrimination, does AI technology exacerbate inequities that already exist? In order to discuss an answer to such a question, a basis of understanding of racism in policing, regardless of the impact of AI, must be established. According to the Prison Policy Initiative, Black people have contact with the police through traffic stops more than other racial groups. In addition, 9% of Black drivers were searched or arrested during traffic stops in 2022, as compared to 4% of Latin(x) drivers, 3% of white drivers, and 5% of drivers of other races.¹² Not only are Black Americans at more risk of being targeted by police in supposedly mundane circumstances, like traffic stops, but they are also more likely to be attacked by the police during those circumstances than other racial groups. Per the 2024 Police Violence Report, “Black people were more likely to be killed by police, more likely to be unarmed [when killed], and less likely to be threatening someone when killed.”¹³ Using the years 2018-2024 to exemplify a broader pattern, the 2024 Police Violence Report found that, year after year, Black people in the US are killed at a disproportionately higher rate by police than any other racial group. According to an article published by The Sentencing Project in 2023, communities of color are overpoliced through discriminatory and biased traffic stops, pedestrian searches, and drug-related arrests. Additionally, the Sentencing Project’s One in Five series points to the statistic that Black Americans are imprisoned at five times the rate of white Americans, a stark example of the racially-fueled crisis of mass incarceration in the US.¹⁴ As another example, Sandra Susan Smith writes in Bias, Distrust, and Trauma: Racial Disparities in Boston Residents’ Experiences with Law Enforcement and Related Outcomes that “with few exceptions, Black Bostonians experience disparate treatment by law enforcement within categories of gender, age, educational attainment, neighborhood of residence, and income status.”¹⁵ Thus, even without the discriminatory impacts of artificial intelligence on the criminal justice system, it is abundantly clear that policing is mixed into a web of systemic racism. This foundational understanding is necessary when developing a complete understanding of the impacts of AI in criminal justice. AI is not a standalone source of discrimination, but rather a compounding agent of harm.
The National Association for the Advancement of Colored People (NAACP) has published an issue brief to their website under the title Artificial Intelligence in Predictive Policing.¹⁶ The article aims to provide valuable information on the background of the use of AI by law enforcement and the challenges AI presents to the Black community, and it also provides recommendations for policies which could address such hurdles. The NAACP specifically calls upon state legislators to evaluate the use of predictive policing, as well as to regulate AI usage within law enforcement agencies. As the NAACP points out, there is a growing mountain of evidence that points to how AI technology, as efficient as it is, is concerningly subjective regarding racial biases. Since predictive policing is marked by the use of data to forewarn law enforcement of indicators of criminal activity, artificial intelligence in predictive policing can provide law enforcement with information on various details, including but not limited to the following: the identification of unknown individuals; the location of an individual; the analysis and management of evidence; and the discernment of future crimes, perpetrators, and victims.¹⁷ In Artificial Intelligence in Predictive Policing, the NAACP urges its readers to consider how the data being utilized to fuel decision-making in policing comes from historical criminal data and police activity. Given the US’ troubling history of racial profiling and other forms of systemic racism in the world of criminal justice, it is disconcerting that this biased data is what is being fed to AI models to establish their library of data.
Already disproportionately impacted by overpolicing, the Black community faces the compounding threat of discriminatory criminal laws,¹⁸ not to mention the separate but terrifying potential threat of police brutality. The NAACP makes it clear that the negative impacts of law enforcement on communities of color, in particular the Black community, include challenges such as bias and discrimination, lack of transparency, and the loss of trust. It includes a note that AI models are able to inherit the biases woven into historical crime data, which fuels discrimination in policing practices. Also, the data fed to AI software does not allow for general public input or commentary, which limits the ability of the software to accurately ‘understand’ how decisions are made with respect to policing. Given that overpolicing has already done tremendous damage to the Black community, it is both unsurprising and understandable that many Black Americans have a tense relationship with the police. The NAACP points out that decisions made by law enforcement that are based on flawed AI can only exacerbate tensions based on an extensive history of racial discrimination. Beyond the discussion of the challenges facing communities of color, in particular the Black community, the NAACP’s issue brief provides a list of recommendations for how the criminal justice system can effectively manage AI risks in predictive policing. From the implementation of rigorous oversight, which would provide objective data from the monitoring of AI usage in policing, to completely banning the use of biased historical crime data to inform AI models, the NAACP provides a series of proposals that would make AI in policing more objective and equitable for Americans. Implementing any of these recommendations, even if not all of them, would yield a palpable difference in police activities by setting strict standards of conduct that enforce humanity and accountability in how law enforcement bodies treat potential offenders.
Artificial intelligence disproportionately hurts communities of color not just in policing, but also in the courts and prisons. Harvard Law School’s Center on the Legal Profession published an article written by Richard Hua in March 2025 that centered on AI and racial bias in legal decision-making. In the article, Hua explores the extent to which AI technological decision-making can be influenced by racial bias as compared to the decisions made by human judges in the American court system.¹⁹ Utilizing multi-variable analysis of legal decisions, Hua determines that, while AI can improve efficiency and consistency in legal and judicial processes, it is also susceptible to racial bias. In another article published by Tulane University, Keith Brannon analyzes the reliance on AI software by judges in the criminal sentencing process, noting that, in a large number of cases, judges that relied on AI were prone to discriminatory decision-making when it came to adjudicating the sentences of Black offenders, despite the AI model’s promise of objectivity.²⁰ To elaborate upon this point, Brannon notes that the AI model improved objectivity when taking into account that judges tend to be more lenient with female offenders than with males. While the AI model seemed to correct this issue, it did not do the same for racial bias in sentencing, where, instead, the AI model perpetuated existing biases that resulted in Black offenders having longer average sentences than their white counterparts. Brannon references a 2023 study in this article, titled AI Enforcement: Examining the Impact of AI on Judicial Fairness and Public Safety, which confirms that judges’ discretion regarding AI software’s recommendations favoring white offenders is a danger to public safety.²¹ The study notes evidence indicating a racial bias that favors white offenders over Black offenders. From these resources, it is apparent that AI’s discrimination towards non-white offenders in the American court system places communities of color, specifically Black communities, at a disturbingly unfair disadvantage in sentencing.
In 2021, Xinxun Li published Analysis of Racial Discrimination in Artificial Intelligence from the Perspective of Social Media, Search Engines, and Future Crime Prediction Systems.²² In this article, Li argues that the achievement of racial equality is unlikely to be achieved in part due to the fact that humans are responsible for the inputting of information into AI software. If any of the programmers have any racial biases, Li makes the point that it will show in the AI’s output. In combination with the use of historical data mentioned previously, this article presents an additional layer to how artificial intelligence can be discriminatorily biased from its inception. Furthermore, Li creates an example of the COMPAS artificial intelligence algorithm, which rated an African American woman who had stolen a child’s scooter and bicycle as a high risk for future crime. Despite the algorithm's prediction, she did not in fact commit later offenses, while a white man deemed low-risk by the algorithm went on to commit additional crimes. The COMPAS algorithm can thus be seen as a warning for the power of AI technology to discriminate based on race, as “Black defendants were still 77 percent more likely to be pegged as at higher risk of committing a future violent crime and 45 percent more likely to be predicted to commit a future crime of any kind.”²³ Li concludes his report by recommending that, in order for racial biases in AI models to be eliminated and for racial discrimination to be reduced, American society must evaluate and recognize the origins of racial bias.
To fully illustrate that artificial intelligence’s discriminatory biases impact the criminal justice system in ways that are incredibly harmful, it is essential to discuss the role that AI models play outside of prisons themselves—i.e., what influence does artificial intelligence software have in terms of rehabilitation, including the reintegration of formerly incarcerated individuals into society following the end of their sentences? For one, AI can disproportionately impact tenant screening services for rented property, e.g., apartment buildings or condominiums. The Georgetown Journal on Poverty Law & Policy explains that tenant screening scores create the illusion of objectivity and are asserted by landlords as effective and efficient tools to rank rental applicants without bias.²⁴ However, evidence points to the fact that these programs, responsible for checking applicants’ credit scores, prior eviction history, and criminal background, are regularly responsible for returning “incorrect, outdated, or misleading information that landlords use to disproportionately deny applications to Black and Latino renters.” Through this evidence, it becomes overtly clear that AI does not typically encourage positive outcomes related to tenancy; but one’s ability to rent an apartment or condominium is not where the biases end.
As was eloquently phrased by the National Homelessness Law Center (NHLC): “Poverty, homelessness, and mass incarceration in the United States cannot be decoupled from the legacy of centuries of systemic racism.”²⁵ The NHLC writes that homelessness feeds directly into the criminal legal system, and the criminal legal system feeds back into homelessness; this feedback loop presents a particularly salient issue for communities of color such as Black communities, since Black people make up one third of those in the prison population but less than one fifth (13%) of the general population. Artificial intelligence plays a role here, too. According to the University of Washington, AI technology reveals biases in ranking the names of job applicants based on how the technology perceives the race and gender of the applicant.²⁶ If Black individuals are already at a disproportionate risk of ending up in the prison system, which already makes it more difficult for someone to find housing after being released into general society, and that same individual could be racially profiled on a job application and subsequently not hired, such a person would be at an increased risk of facing housing instability. With artificial intelligence influencing disparities in tenancy and employment, it is reasonable to draw the conclusion that, in addition to the impact it has on policing and sentencing, artificial intelligence models have the capacity to have a disproportionately negative impact on communities of color, especially Black communities, when it comes to homelessness.
Despite the argument that AI models bring a discriminatory bias to the American criminal justice system, that does not mean that there are no benefits to utilizing this emerging technology. AI has the potential to improve efficiency in judicial processes and reduce the capacity for human error in legal filing activities. For example, in regard to the increased efficiency of legal writing and creation of courthouse documentation, Thomson Reuters cites Miami-Dade County Public Defender Carlos Martinez as endorsing large language model artificial intelligence as a means to assist the public defenders’ office in the initial drafting process for memorandums and other document preparation.²⁷ The Miami-Dade County Public Defender’s Office—among the first in the nation to begin utilizing AI technology in this manner—maintains that AI has the ability to save time in legal processing, which can then, in turn, enhance public defenders’ preparation for court and comprehension of complex and varied caseloads.²⁸ This potential benefit of using new technology could mean a faster and more just system, where those accused of crime are not waiting lengthy periods of time awaiting trial. As AI models consistently and constantly improve, they present even more opportunities for change, either positive or negative, within the criminal justice system. Given the heavy investment that the US’ economic markets have made in the development of these systems, with a staggering total of $252.3 billion in corporate investment in 2024—a 26% increase from the previous year—it is abundantly clear that AI is only going to continue to permeate various societal processes.²⁹ There is an argument to be made that the American criminal justice system, which is archaic and outdated on a variety of levels, needs to modernize, and AI might be just the solution. Since these models are evolving at an exponential rate, it is feasible to improve them so as to address their inherent biases whilst simultaneously retaining their ability to enhance efficiency and effectiveness.³⁰ Awareness of the pitfalls of these models has led to the rapid development of models that can be used for good within the system, yielding potential long-term progress. However, while artificial intelligence does have advantages in terms of saving time, its relative newness and constant need for improvement plays directly into AI’s disproportionately negative role in criminal justice—this technology is too new and unwieldy to be making decisions that can permanently alter the life of a human being. Without proper regulation, which is an arduous process in and of itself, these mechanisms are simply not advanced enough to impact the way that the US polices, arrests, tries, and convicts its citizens. The concept of efficiency is baked into the nation’s socioeconomic structures—it is part of what makes the AI boom so massive in its influence. However, until it is advanced enough to guarantee impartiality and regulated accordingly, it is a hindrance to the basic human rights of many Americans.
Eliezer Yudkowsky, a prominent artificial intelligence researcher, co-founder of the Machine Intelligence Research Institute, and member of TIME’s 2023 list of the “100 Most Influential People in AI,” eloquently summarizes the core message of this essay: “By far the greatest danger of artificial intelligence is that people conclude too early that they understand it.”³¹ The US criminal justice system is a monolith, and one that has been reprimanded countless times since its very conception for serving the privileged few over the overburdened and underserved masses. By weaving AI models into how the US approaches criminal and judicial processes, the country opens itself to compounding the amount of racial bias, and thus discriminatory practice, that already harms vulnerable communities. Until AI advances to the point that it does not pose risks to communities of color and other historically disadvantaged groups, it is a hindrance to the elimination of racism and racialized prejudice that the American system needs to progress.
Glossary
Adjudicate - To settle judicially (in a court of law).
Amalgamation - The combination of multiple distinct elements.
Apartheid - Originally, the systemic oppression of Black South Africans by white South Africans; generally, the institutionalized oppression of one race or ethnic group by another.
Arduous - Difficult or otherwise burdensome.
Artificial intelligence - “A form of technology that enables computers and machines to simulate human learning, comprehension, problem solving, decision making, creativity and autonomy.”
Burgeoning - Something that is growing or expanding.
Discernment - Ability to clearly make decisions or judgements.
Discourse - Conversation, discussion.
Disproportionate - To a larger extent than something else (comparative).
Egalitarian economy - The economic and philosophical concept of levelling the metaphorical playing field by prioritizing equal rights and access for all regardless of socioeconomic standing.
Exacerbate - To a negative, to multiply or make worse.
Exodus - Mass movement of a group of beings collectively exiting one place in pursuit of another.
Hinder - To impede or stand in the way of.
Impetus - Motivation, driving force.
Inherent biases - Biases toward or against certain groups or individuals based on one’s own lived experiences or from exposure to stereotypes.
Latin(x) - An individual or group of individuals with Latin American heritage (Latino/Latina).
Mistretta v. United States - A l989 landmark Supreme Court case that established the constitutionality of the Sentencing Commission, the independent agency established in 1984 to reduce discrepancies in judicial sentencing. The Commission’s guidelines notably did not include rehabilitation as a sentencing option for convicted individuals.
Monolith - Any structure, physical or otherwise, that is deemed immovable or imposing.
Mundane - Seemingly ordinary or plain.
Notion - Concept, idea.
Palpable - An intangible concept or item that feels real enough to be held or seen.
Predictive policing - The use of computer systems to “analyze large sets of data, including historical crime data, to help where to deploy police or to identify individuals who are purportedly more likely to commit or be a victim of crime.”
Punitive - With attention to punishment, often in legal contexts.
Rehabilitation - In the context of criminal justice, the process of reforming offenders by addressing the issues that contribute to their criminal behavior, as opposed to punitive treatment. Historically, criminal rehabilitation served as a guiding principle in the early American prison system that sought to separate offenders from negative influences.
Restitutive justice - A rights-based approach to the theory of justice that views crime as an offense of one individual against the rights of another, which then calls for restitution from the offending individual to the harmed one. For example, restitution in criminal justice can look like an offender financially reimbursing a victim for a financial loss.
Sociology(-ist) - The study of (or, one who studies) society and human interaction.
Solitary confinement - The act of separating an individual from the larger incarcerated population of a jail or prison for a length of time.
Stigma - “A set of negative and often unfair beliefs that a society or group of people have about something.”
Susceptible - To be more likely to be harmed or impacted negatively by something.
Systemic racism - The oppression of a racial group to the advantage of another that is perpetuated by inequity across integrated systems.
‘Use of force’ - The amount of effort required by law enforcement to compel compliance by an unwilling subject.
Footnotes
Gibbs, J. C., Criminal Justice in U.S. History (EBSCO Research Starters, 2022). Retrieved from https://www.ebsco.com/research-starters/history/criminal-justice-us-history.
Warnes, K. (2024). Prison Reform, EBSCO Research Starters, Politics & Government. Retrieved from https://www.ebsco.com/research-starters/politics-and-government/prison-reform.
Chlup, Dominique T. “Chronology of Corrections Education.” Focus on Basics, vol. 7, issue D, Sept. 2005, National Center for the Study of Adult Learning and Literacy, https://www.ncsall.net/index.php@id=865.html.
Leif Wenar, “John Rawls,” Stanford Encyclopedia of Philosophy, Spring 2024 Edition. Retrieved from https://plato.stanford.edu/entries/rawls/.
Robert Martinson, “What Works? Questions and Answers About Prison Reform” (University of Minnesota Duluth [online copy], 1974). Retrieved from https://www.d.umn.edu/~jmaahs/Correctional%20Continuum/Online%20Readings/martinson.pdf.
Prison Policy Initiative, The Debate on Rehabilitating Criminals: Is It True That Nothing Works? March 1989. Retrieved from https://www.prisonpolicy.org/scans/rehab.html.
Office of the United Nations High Commissioner for Human Rights, “Report of the International Independent Expert Mechanism to Advance Racial Justice and Equality (IIEM) on systemic racism, racial discrimination, xenophobia and related intolerance,” report, 2023 (A/HRC/54/CRP.7). Retrieved from https://www.ohchr.org/en/documents/country-reports/ahrc54crp7-international-independent-expert-mechanism-advance-racial.
OHCHR, IIEM Report on Systemic Racism (2023).
OHCHR, “Systemic Racism Pervades US Police and Justice Systems, UN Mechanism Says in New Report Urging Reform,” press release, September 28, 2023. Retrieved from https://www.ohchr.org/en/press-releases/2023/09/systemic-racism-pervades-us-police-and-justice-systems-un-mechanism-racial.
OHCHR, “Systemic Racism Pervades US Police and Justice Systems, UN Mechanism Says,” press release, September 28, 2023. Retrieved from https://www.ohchr.org/en/press-releases/2023/09/systemic-racism-pervades-us-police-and-justice-systems-un-mechanism-racial.
Office of the High Commissioner for Human Rights (OHCHR), “Racism and AI: ‘Bias from the Past Leads to Bias in the Future’,” OHCHR, July 30, 2024. Retrieved from https://www.ohchr.org/en/stories/2024/07/racism-and-ai-bias-past-leads-bias-future.
Scott Prell, “Despite Fewer People Experiencing Police Contact, Racial Disparities in Arrests, Police Misconduct, and Police Use of Force Continue,” Prison Policy Initiative (blog), December 19, 2024. Retrieved from https://www.prisonpolicy.org/blog/2024/12/19/policing_survey_2022/.
Mapping Police Violence, 2024 Police Violence Report (Mapping Police Violence, 2025). Retrieved from https://policeviolencereport.org.
Nazgol Ghandnoosh, One in Five: Racial Disparity in Imprisonment — Causes and Remedies (The Sentencing Project, 2023). Retrieved from https://www.sentencingproject.org/reports/one-in-five-racial-disparity-in-imprisonment-causes-and-remedies/.
Prison Policy Initiative, “Racial and Ethnic Disparities,” (Prison Policy Initiative), N.d., retrieved from https://www.prisonpolicy.org/research/racial_and_ethnic_disparities/. Date accessed: September 10, 2025.
NAACP, Artificial Intelligence in Predictive Policing: Issue Brief (NAACP, 2024). Retrieved from https://naacp.org/resources/artificial-intelligence-predictive-policing-issue-brief.
The Policing Project, “How Policing Agencies Use AI,” Policing Project (AI Explained), September 6, 2024. Retrieved from https://www.policingproject.org/ai-explained-articles/2024/9/6/how-policing-agencies-use-ai.
Elizabeth Hinton, LeShae Henderson & Cindy Reed, An Unjust Burden: The Disparate Treatment of Black Americans in the Criminal Justice System (Vera Institute of Justice, May 2018). Retrieved from https://vera-institute.files.svdcdn.com/production/downloads/publications/for-the‑record-unjust-burden-racial-disparities.pdf.
Richard Hua, “AI and Racial Bias in Legal Decision‑Making: A Student Fellow Project,” Harvard Law School Center on the Legal Profession (Insight), March 18, 2025. Retrieved from https://clp.law.harvard.edu/knowledge-hub/insights/ai-and-racial-bias-in-legal-decision-making-a-student-fellow-project/.
Keith Brannon, “AI Sentencing Cut Jail Time for Low‑Risk Offenders, but Study Finds Racial Bias Persisted,” Tulane University News, January 23, 2024. Retrieved from https://news.tulane.edu/pr/ai-sentencing-cut-jail-time-low-risk-offenders-study-finds-racial-bias-persisted.
Yi‑Jen (Ian) Ho, Wael Jabr & Yifan Zhang, AI Enforcement: Examining the Impact of AI on Judicial Fairness and Public Safety (SSRN, 2024). Retrieved from https://ssrn.com/abstract=4533047.
Xinxun Li, “Analysis of Racial Discrimination in Artificial Intelligence from the Perspective of Social Media, Search Engines, and Future Crime Prediction Systems,” in Proceedings of the 6th International Conference on Contemporary Education, Social Sciences and Humanities (ICCESSH 2021), Advances in Social Science, Education and Humanities Research, vol. 575 (Atlantis Press, 2021), 172–177. DOI: 10.2991/assehr.k.210902.029. Retrieved from https://www.atlantis-press.com/proceedings/iccessh-21/125960619.
Julia Angwin, Jeff Larson, Surya Mattu & Lauren Kirchner, “Machine Bias: There’s Software Used across the Country to Predict Future Criminals — And It’s Biased Against Blacks,” ProPublica, May 23, 2016. Retrieved from https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing.
Lauren Karpinski, “The Discriminatory Impacts of AI-Powered Tenant Screening Programs,” Georgetown Journal on Poverty Law & Policy (blog), July 12, 2025. Retrieved from https://www.law.georgetown.edu/poverty-journal/blog/the-discriminatory-impacts-of-ai-powered-tenant-screening-programs/.
National Homelessness Law Center, Racism, Homelessness, and the Criminal & Juvenile Legal Systems (Washington, DC: National Homelessness Law Center, August 2020), pp. 1–8. Retrieved from https://homelesslaw.org/wp-content/uploads/2020/08/Racism-Homelessness-and-Criminal-Legal-Systems.pdf
Stefan Milne, “AI tools show biases in ranking job applicants’ names according to perceived race and gender,” UW News, October 31, 2024. Retrieved from https://www.washington.edu/news/2024/10/31/ai-bias-resume-screening-race-gender/.
Allyson Brunette, “Humanizing Justice: The Transformational Impact of AI in Courts, from Filing to Sentencing,” Thomson Reuters Institute, October 25, 2024. Retrieved from https://www.thomsonreuters.com/en-us/posts/ai-in-courts/humanizing-justice/.
Rabihah Butler, “Revolutionizing Rights: Spotlight on Carlos Martinez, Esq.,” Thomson Reuters Institute, September 12, 2024. Retrieved from https://www.thomsonreuters.com/en-us/posts/government/revolutionizing-rights-carlos-martinez/.
Stanford HAI, Economy, in The 2025 AI Index Report (Stanford, CA: Human-Centered AI Institute), 2025. Retrieved from https://hai.stanford.edu/ai-index/2025-ai-index-report/economy.
Adam Zewe, “Researchers Reduce Bias in AI Models While Preserving or Improving Accuracy,” MIT News, December 11, 2024. Retrieved from https://news.mit.edu/2024/researchers-reduce-bias-ai-models-while-preserving-improving-accuracy-1211.
Eliezer Yudkowsky, Artificial Intelligence as a Positive and Negative Factor in Global Risk, in Global Catastrophic Risks, ed. Nick Bostrom & Milan M. Ćirković (New York: Oxford University Press, 2008), pp. 308–345. Retrieved from https://www.semanticscholar.org/paper/Artificial-Intelligence-as-a-Positive-and-Negative-Yudkowsky/fcf7368061a544a09d16826eb4c5a8463ee5482e.
