Predictive policing, risk assessment, sentencing assistance.
Definition & scope
The cross-jurisdiction picture below shows how each of 45 tracked instruments treats this topic. The patterns vary substantially — and 39 regimes are silent, leaving gaps that future policy work could address.
Regulatory approaches
The three instruments that govern this topic do so through markedly different modalities, which the coverage table records as verdicts but does not unpack. The EU AI Act operates as tiered product regulation. It first draws an absolute red line: Article 5(1)(d) prohibits AI that assesses or predicts an individual's risk of committing a criminal offence "based solely on the profiling of a natural person or on assessing their personality traits and characteristics," a ban applicable since 2 February 2025 (Reg. (EU) 2024/1689, Art. 5(1)(d), Art. 113(a)). Tools that merely support a human assessment already grounded in "objective and verifiable facts directly linked to a criminal activity" fall outside the ban (Art. 5(1)(d)). Everything else in this domain — recidivism risk scoring, polygraph-type systems, evidence-reliability evaluation, crime analytics — is classed high-risk under Annex III point 6, triggering conformity assessment, logging, human oversight, and, for public-authority deployers, an ex ante Fundamental Rights Impact Assessment (Art. 27(1)); high-risk obligations apply from 2 August 2026 (Art. 113). Even where the Act is silent on a specific technique such as facial recognition, scholars note the surrounding EU framework already supplies norms "directly or indirectly applicable" to its law-enforcement use 1. By contrast, US Executive Order 14110 used a soft, study-first mechanism: §7.1(b) directed the Attorney General to report on AI in the criminal-justice system rather than impose binding controls (EO 14110, §7.1(b)). The Council of Europe Framework Convention is rights-based and outcome-oriented, requiring effective remedies (Art. 14) and procedural safeguards including notice and the ability to contest AI-informed decisions (Art. 15), but leaving implementation to each Party — a discretion that, on the evidence of national case studies, tends to produce fragmented governance and weak transparency over which tools are actually deployed 2.
Key fault lines
Beyond the empirical disputes the article already catalogues, jurisdictions diverge sharply on regulatory design. The first fault line is prohibition versus permission. The EU treats one application — purely profiling-based individual crime prediction — as an unacceptable risk warranting an outright ban (Reg. (EU) 2024/1689, Art. 5(1)(d)), whereas no US federal instrument bans any criminal-justice AI use; EO 14110 commissioned study rather than restriction (EO 14110, §7.1(b)). Commentators dispute whether the EU ban bites at all, since the "solely"-profiling threshold and the carve-out for tools supporting fact-based human assessment may leave most deployed predictive-policing systems untouched (Free, European Law Blog, 2024; Future of Privacy Forum analysis, 2026). That contest is sharpened by evidence that claimed effectiveness — not just rights concerns — is what underpins the political legitimacy of predictive policing in the UK and US, even as systematic review flags persistent algorithmic bias and data-concentration worries 3. A broader synthesis of a decade of research frames the cross-cutting stakes as "algorithmic bias, opacity, and due process," recommending equity and accountability safeguards that map onto exactly these design choices 4. A second fault line concerns scope exclusions. The Council of Europe Convention, the only binding treaty here, exempts national-security and defence activities and lets each Party choose whether to apply its rules to private actors at all — an opt-in for the private sector that civil-society groups and the European Data Protection Supervisor warned could hollow out protection and shelter state surveillance (CETaS/Alan Turing Institute, 2024; CAIDP treaty brief, 2025). A third fault line is durability: the US position is contested even domestically, EO 14110 having been revoked on 20 January 2025 (EO 14148, Fed. Reg. 2025-01901), illustrating that executive-action governance of policing AI is reversible in a way statute and treaty are not.
Trajectory / what's changing
The governance picture for this topic is in active flux on a dated timeline. The EU AI Act's prohibition on solely-profiling individual crime prediction became applicable on 2 February 2025, the first hard legal constraint specific to predictive policing anywhere (Reg. (EU) 2024/1689, Art. 5(1)(d); Art. 113(a)); the high-risk obligations governing recidivism and law-enforcement risk tools under Annex III point 6 — conformity assessment, logging, human oversight, and the Article 27 fundamental-rights impact assessment for public-authority deployers — follow on 2 August 2026 (Art. 113). These constraints respond to documented failure modes that statute is meant to discipline: feedback loops that send police repeatedly to the same neighbourhoods regardless of the true crime rate (Ensign et al., 2018; FRA, Bias in algorithms, 2022), and formal impossibility results showing a risk score cannot be simultaneously calibrated and equalised across groups 56 — making the choice between an outright ban and high-risk obligations a genuinely contested design call rather than a technicality. In the United States the direction reversed: Executive Order 14110, whose §7.1(b) had tasked the Attorney General with reporting on criminal-justice AI, was revoked on 20 January 2025 by Executive Order 14148, and the successor Executive Order 14179 (23 January 2025) reorients federal policy toward removing barriers to AI rather than constraining its criminal-justice use — leaving the US federal layer effectively silent on this topic, as the coverage map now records. Internationally, the Council of Europe Framework Convention opened for signature on 5 September 2024 and had drawn dozens of signatories within months (Council of Europe, CETS No. 225), but it confers no obligations until ratified and domestically implemented, so its Article 14–15 remedies and safeguards remain prospective rather than operative for affected individuals.
Coverage across jurisdictions
Historical primacy & cross-jurisdiction tension
First addressed by UNESCO Recommendation on the Ethics of Artificial Intelligence on (implicit). Subsequent regimes have either codified, diverged from, or remained silent on this baseline.
- Forum-shoppingEU AI Act↔Executive Order 14179 — Removing Barriers to American Leadership in AI
- Forum-shoppingExecutive Order 14110 on Safe, Secure, Trustworthy AI↔Interim Measures for Generative AI Service Management
- Forum-shoppingCouncil of Europe Framework Convention on AI↔G7 Hiroshima AI Process Code of Conduct
Compare jurisdictions: EU vs US · EU vs UK · EU vs CN
Enforcement & impact
Silent regimes — gap signal
Instruments that do not address AI in Criminal Justice — candidates for future policy work.
- Executive Order 14179 — Removing Barriers to American Leadership in AIUS
- Interim Measures for Generative AI Service ManagementCN
- G7 Hiroshima AI Process Code of ConductG7
- OECD AI Principles (Recommendation)OECD
- UN GA Resolution on Safe, Secure, Trustworthy AIUN
- NIST AI Risk Management FrameworkUS
- Bletchley Declaration on AI Safetyglobal
- Seoul Declaration on Safe, Innovative and Inclusive AIglobal
- NIST AI RMF Generative AI ProfileUS
- California SB-1047: Safe and Secure Innovation for Frontier AI Models ActUS
- India Digital Personal Data Protection Act + AI Advisory (MEITY)IN
- Brazil AI Bill (PL 2338/2023)BR
- ASEAN Guide on AI Governance and EthicsASEAN
- African Union Continental AI StrategyAfrican_Union
- Anthropic Responsible Scaling Policy (RSP) v2US
- OpenAI Preparedness FrameworkUS
- Google DeepMind Frontier Safety FrameworkUS
- Meta Frontier AI FrameworkUS
- UK-US AI Safety Institute Memorandum of Understandingglobal
- White House Voluntary AI CommitmentsUS
- Singapore Model AI Governance Framework for Generative AISG
- Japan METI AI Guidelines for BusinessJP
- General Data Protection Regulation (GDPR)EU
- EU General-Purpose AI Code of PracticeEU
- OMB Memorandum M-24-10 (Advancing Governance, Innovation, and Risk Management for Agency Use of AI)US
- GSA Generative AI and Specialized Computing Infrastructure Acquisition Resource GuideUS
- DoD Responsible AI Strategy and Implementation PathwayUS
- FedRAMP AI Cloud Procurement GuidanceUS
- DFARS Subpart 252.204 (Safeguarding Covered Defense Information and Cyber Incident Reporting)US
- California SB-53: Transparency in Frontier Artificial Intelligence Act (TFAIA)US
- California SB 243: Companion ChatbotsUS
- California SB 942: AI Transparency ActUS
- Revised Product Liability Directive (Directive (EU) 2024/2853)EU
- Directive (EU) 2024/2831 on improving working conditions in platform workEU
- Provisions on the Administration of Deep Synthesis of Internet Information ServicesCN
- New York RAISE Act: Responsible AI Safety and Education ActUS
- TAKE IT DOWN Act (Tools to Address Known Exploitation by Immobilizing Technological Deepfakes on Websites and Networks Act)US
- Japan AI Promotion Act (Act on the Promotion of Research, Development and Utilization of AI-Related Technologies)JP
- UN Global Digital CompactUN
See also
Further reading
15 academic & grey-literature sources bearing on this topic — catalogued metadata with a primary link; one-line findings are ✦ AI-generated summaries, labeled as such (charter §7.9). Browse the full literature index.
- Machines of justice: A systematic review of AI applications in policing and criminal justice Peer-reviewed✦ AISynthesises a decade of AI-in-criminal-justice research, flagging "algorithmic bias, opacity, and due process" and recommending safeguards for equity and accountability.
- The Effectiveness of Big Data-Driven Predictive Policing: Systematic Review Peer-reviewed✦ AISystematic review of 161 articles finds claimed effectiveness underpins legitimacy of predictive policing in the UK and US while algorithmic bias and data-concentration concerns persist.
- The Use of Facial Recognition Technology by Law Enforcement in Europe: a Non-Orwellian Draft Proposal Peer-reviewed✦ AIArgues the EU framework already contains norms "directly or indirectly applicable to facial recognition" in policing, and drafts a dedicated rights-protective law for its use.
- Transparency, Governance and Regulation of Algorithmic Tools Deployed in the Criminal Justice System: a UK Case Study Peer-reviewed✦ AIUK case study maps algorithmic tools used across the criminal-justice system and finds fragmented governance and weak transparency over their deployment.
- The accuracy, fairness, and limits of predicting recidivism Peer-reviewed✦ AIFinds COMPAS "is no more accurate or fair than predictions made by people with little or no criminal justice expertise"; a two-feature linear model matches it.
- Fairness in Criminal Justice Risk Assessments: The State of the Art Peer-reviewed✦ AISurveys six fairness definitions: "impossible to maximize accuracy and fairness at the same time, and impossible simultaneously to satisfy all kinds of fairness".
- Runaway Feedback Loops in Predictive Policing Peer-reviewed✦ AIProves mathematically that learning from discovered-crime data sends police repeatedly to the same neighbourhoods "regardless of the true crime rate," and shows how to correct it.
- Assessing Risk Assessment in Action Peer-reviewed✦ AIEmpirical study of Kentucky's mandatory pretrial risk assessment finds an initial small detention drop that dissipated as judges reverted, with limited net change and modest disparity effects.
- Fair Prediction with Disparate Impact: A Study of Bias in Recidivism Prediction Instruments Peer-reviewed✦ AIShows a recidivism instrument satisfying predictive parity "may lead to considerable disparate impact when recidivism prevalence differs across groups".
- Inherent Trade-Offs in the Fair Determination of Risk Scores Peer-reviewed✦ AIProves calibration and balanced error rates cannot coexist: "except in highly constrained special cases, there is no method that can satisfy these three conditions simultaneously".
- Disparate Impact in Big Data Policing Peer-reviewed✦ AIArgues data-driven predictive policing can produce disparate racial impacts even when well-intentioned, and proposes algorithmic impact statements as a legal remedy.
- Randomized Controlled Field Trials of Predictive Policing Peer-reviewed✦ AIFirst RCT field trials of predictive policing report algorithmic hotspot predictions led to crime reductions versus analyst-designated patrols.
- Bias in algorithms - Artificial intelligence and discrimination Official (grey)✦ AIEU agency report whose predictive-policing feedback-loop simulation shows biased crime data amplifying over-policing of minorities.
- Predictive Policing: The Role of Crime Forecasting in Law Enforcement Operations Research institute✦ AIFoundational study framing four predictive-policing method families; cautions the tools forecast risk, not events.
- Evaluating Algorithmic Risk Assessment Peer-reviewed✦ AICross-jurisdiction legal evaluation of pretrial algorithmic risk-assessment tools and their contested fairness and accuracy.
References
Sources cited inline in the analysis (linked from the superscript markers), then the primary instrument sources behind the classifications.
- Raposo (2023) The Use of Facial Recognition Technology by Law Enforcement in Europe: a Non-Orwellian Draft Proposal, European Journal on Criminal Policy and Research. 10.1007/s10610-022-09512-y — Argues the EU framework already contains norms "directly or indirectly applicable to facial recognition" in policing, and drafts a dedicated rights-protective law for its use. ↩
- Miri Zilka, Holli Sargeant, Adrian Weller (2022) Transparency, Governance and Regulation of Algorithmic Tools Deployed in the Criminal Justice System: a UK Case Study, Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, a. 10.1145/3514094.3534200 — UK case study maps algorithmic tools used across the criminal-justice system and finds fragmented governance and weak transparency over their deployment. ↩
- Youngsub Lee, Ben Bradford, Krisztian Posch (2024) The Effectiveness of Big Data-Driven Predictive Policing: Systematic Review, Justice Evaluation Journal. 10.1080/24751979.2024.2371781 — Systematic review of 161 articles finds claimed effectiveness underpins legitimacy of predictive policing in the UK and US while algorithmic bias and data-concentration concerns persist. ↩
- Shai Farber (2026) Machines of justice: A systematic review of AI applications in policing and criminal justice, The Police Journal: Theory, Practice and Principles. 10.1177/0032258X261439572 — Synthesises a decade of AI-in-criminal-justice research, flagging "algorithmic bias, opacity, and due process" and recommending safeguards for equity and accountability. ↩
- Kleinberg, Mullainathan & Raghavan (2017) Inherent Trade-Offs in the Fair Determination of Risk Scores, ITCS 2017. arXiv:1609.05807 — Proves calibration and balanced error rates cannot coexist: "except in highly constrained special cases, there is no method that can satisfy these three conditions simultaneously". ↩
- Chouldechova (2017) Fair Prediction with Disparate Impact: A Study of Bias in Recidivism Prediction Instruments, Big Data. 10.1089/big.2016.0047 — Shows a recidivism instrument satisfying predictive parity "may lead to considerable disparate impact when recidivism prevalence differs across groups". ↩
- EU-AIA-2024: Annex III §6 (high-risk: law enforcement)
- US-EO-14110: §7.1(b) (DOJ AI use review)
- UK-WHITEPAPER-2023: Forensic Information Databases Strategy Board
- COE-AI-CONV: Art. 14 (procedural safeguards)
- UNESCO-AI-ETHICS-2021: Ethical-governance section, paras 62-63 — names law enforcement + the judiciary as sensitive use cases requiring oversight; no dedicated criminal-justice regime
- IT-AILAW-2025: Art. 15 — in judicial use of AI, decisions on legal interpretation/application, evaluation of facts and evidence, and adoption of measures are always reserved to the magistrate; AI limited to organisational/administrative support. Art. 24(2)(h) delegates a future regime for AI in policing.
Cite this article 8 formats · BibTeX, RIS, APA, Chicago, … · 1-click copy
Persistent identifier: https://policywindow.org/wiki/criminal-justice — committed-stable URL with content-versioning via ?asOf= (rollout pending per methodology §7). DOIs via Zenodo are on the roadmap.
Article tools — track changes, suggest an edit
View history — every captured revision of this article · What links here
6 instruments tracked.
Does governance work? — the social-science evidence
What the peer-reviewed social science shows: whether the harm this topic addresses is empirically real, and whether governance of it works. The badge is the epistemic status of the evidence(not the policy debate) — “thin” or “absent” efficacy evidence is itself a finding (the “second silence”). Each epistemic-status label is Policy Window's editorial assessment of the cited evidence base (a structured classification), not a verdict any single source issues.
Whether algorithmic risk assessment reproduces racial disparity is a genuine, partly mathematically irreducible dispute rather than merely an unresolved measurement question. ProPublica's analysis of COMPAS in Broward County found Black defendants who did not reoffend were nearly twice as likely to be flagged high-risk as comparable white defendants (44.9% vs 23.5% false-positive rate; Angwin et al. 2016), and Dressel & Farid (2018) showed COMPAS is no more accurate (65.2%) than untrained laypeople (67.0%); the developer's reanalysis (Flores, Bechtel & Lowenkamp 2016) found the same tool satisfies predictive parity and calibration across race. Honest caveat: Chouldechova (2017) proved both sides can be correct simultaneously — when recidivism base rates differ across groups, equal calibration and equal error rates cannot both hold, so the disagreement is partly definitional, not merely a data dispute to be settled.
Sources: Angwin, Larson, Mattu & Kirchner 2016 (ProPublica, 'Machine Bias'); Dressel & Farid 2018 (Science Advances 4:eaao5580); Flores, Bechtel & Lowenkamp 2016 (Federal Probation 80(2):38); Chouldechova 2017 (Big Data 5(2):153)
Rigorous evidence that governing criminal-justice algorithms — mandating, auditing, or adopting risk tools — reduces the racial-disparity harm that motivates the rules is essentially absent. The leading real-world impact evaluation, Stevenson's (2018) study of Kentucky's mandatory pretrial risk-assessment law (>1M cases), found only a small increase in pretrial release that eroded as judges reverted to prior habits, with no reduction in racial disparities in pretrial detention. The closest analogue evaluations measure operational crime outcomes, not equity, and are largely null: Chicago's Strategic Subjects List had no effect on victimization (Saunders, Hunt & Hollywood 2016) and the only randomized predictive-policing trials tested crime reduction, not disparate impact (Mohler et al. 2015) — so the evidence that any governance regime measurably reduces algorithmic racial disparity is itself missing.
Sources: Stevenson 2018 (Minnesota Law Review 103:303); Saunders, Hunt & Hollywood 2016 (Journal of Experimental Criminology 12(3):347); Mohler et al. 2015 (JASA 110(512):1399)