Policing Future Risk in Australia: The NSW Suspect Targeting Management Plan, Operation Tepito, and Successor Practices

Policing Future Risk in Australia: The NSW Suspect Targeting Management Plan, Operation Tepito, and Successor Practices

Executive assessment and research method

The New South Wales Police Force’s Suspect Targeting Management Plan was one of Australia’s longest-running person-focused preventive-policing systems. It operated in successive forms for roughly twenty-three years, beginning in or around 2000 and ending for children in 2023 and for adults by December 2023. Its stated objective was to reduce recidivism by identifying people assessed as likely to offend again, placing them under individualized management plans, and directing police to give them concentrated “proactive” attention. That attention could include observation, intelligence collection, stops and searches, home visits, bail checks, consorting warnings, enforcement of firearms restrictions, arrest, diversionary engagement, referrals, and coordination with other agencies. The plan itself supplied no new legal powers; every coercive encounter still had to satisfy ordinary statutory or common-law requirements. citeturn0search5turn12view0turn19view0

The program changed significantly over time. The earlier STMP-II used an automatically calculated risk-assessment instrument that combined judgments about the local “crime environment” with “person disposition” variables such as offending volume, prior imprisonment and other indicators of recidivism. It generated classifications described in publicly obtained material as extreme, high, medium or low risk. Yet the calculation occurred only after local police had already identified someone as a potentially suitable target, and the inputs themselves included human judgments. STMP-III, introduced in November 2020, replaced that instrument with a Crime Severity Index, a comprehensive profile prepared through the Chimera intelligence platform, a written nomination justification, senior-officer approval and a Target Action Plan. The Crime Severity Index was not mandatory or determinative: officers could nominate people with very low or zero scores, and selection ultimately remained an exercise of police discretion. citeturn24search4turn24search11turn12view2turn12view3

For that reason, STMP is best classified as a hybrid of person-based predictive policing, intelligence-led policing and offender management, organized around a recidivism-control objective. It was predictive because its defining intervention criterion was prospective risk rather than merely investigation of a completed offense; intelligence-led because local intelligence units assembled records, selected nominees and designed responses; and offender-management oriented because named individuals received continuing plans, reviews and coordinated interventions. It was not a purely automated predictive system: neither STMP-II’s score nor STMP-III’s Crime Severity Index autonomously generated a legally binding target list. Nor was it a complete focused-deterrence program of the kind evaluated internationally, because clear deterrence messaging, community legitimacy, social-service commitments and meaningful offers of assistance were inconsistent or peripheral. citeturn24search13turn13view5turn23view2turn9search15

The Law Enforcement Conduct Commission’s Operation Tepito provides the most authoritative account of actual implementation. Its 2020 interim report examined all 429 young people placed on STMP between August 1, 2016 and August 1, 2018. Its 2023 final report examined 133 young people carried into, or newly selected for, STMP-III during its first six months, with police-interaction records reviewed through February 28, 2022. The Commission found that the program’s application to young people remained dominated by intrusive disruption, weak documentation, inadequate consideration of childhood and complex needs, profound Aboriginal overrepresentation, and confusion between the existence of an STMP plan and the lawful grounds required for each stop, search, entry or bail check. It concluded that some interactions were or may have been unlawful. citeturn19view3turn18view3

The Commission expressed the opinion that use of STMP-III on young people during the reviewed period met the statutory threshold for “agency maladministration”: conduct that was or could be unreasonable, unjust, oppressive or improperly discriminatory. It did not, however, make a formal adverse finding after NSW Police discontinued the program. That distinction is important. The formal historical record is not that a court declared STMP as a whole unlawful, nor that LECC made misconduct findings against every officer or every interaction. It is that the specialist integrity body found systemic features capable of constituting agency maladministration, documented interactions that appeared unlawful or insufficiently justified, and treated discontinuation as having removed much of the practical utility of a formal finding. citeturn3view7turn1search5turn16search2

NSW Police disputed parts of the criticism. It maintained that STMP was preventive rather than punitive, that nominees were individually assessed, that alternative programs and positive interventions could be incorporated, and that later reforms strengthened governance, recordkeeping and youth considerations. It also criticized the Youth Justice Coalition’s 2017 report as relying on a small, partly anecdotal group that included suspected rather than confirmed targets and did not always identify which version of STMP applied. LECC answered that its final conclusions rested on police-supplied records for complete or defined cohorts, not on the coalition’s case studies. citeturn19view3turn18view3

The most favorable quantitative evaluation, published by the NSW Bureau of Crime Statistics and Research, found statistically significant associations between STMP-II placement and lower recorded property offending, especially among juveniles, while also finding an increased probability of imprisonment. BOCSAR expressly revised its report to state that the estimates were not causal and corrected an error that had initially obscured the increased custodial-sentence risk for juveniles. Independent methodological critiques argued that the apparent reduction could arise from selection near a temporary peak in recorded offending followed by regression to the mean, and from model and comparison-group choices. BOCSAR defended aspects of its matching and robustness analyses but retained the non-causal qualification. citeturn23view3turn13view7turn1search8turn1search10turn1search12turn1search14

The best-supported overall judgment is therefore asymmetric. STMP plausibly concentrated police resources on people who committed substantial volumes of recorded crime, and recorded offending often fell after selection. The available studies cannot determine how much of that fall was caused by deterrence, incapacitation, natural fluctuation, maturation, concurrent sanctions, changed recording or other interventions. Nor can aggregate crime reductions answer whether repeated contacts were lawful, proportionate, fairly distributed or developmentally appropriate. Effectiveness in reducing recorded incidents, even if ultimately established, would not cure unlawful searches, unauthorized entries, discriminatory selection or procedurally unfair administration.

As of August 2, 2026, STMP itself has not returned. The publicly identified replacement is the Prevention, Disruption, Response process, implemented through revised standard operating procedures from mid-2024. NSW Police says PDR is issue-focused rather than an offender-centric targeting program and that individuals are not “policed pursuant to” PDR. Local commands use crime data, meetings, community engagement and tasking to develop strategic responses. LECC describes PDR as the current local crime-management framework and continues to monitor proactive operations. Public information remains too sparse to determine whether particular commands use informal person lists, how interventions are attributed to PDR, or whether demographic disparities can be audited. NSW Police says it holds no demographic dataset specific to the application of the PDR procedures because ordinary encounters are recorded through existing systems. citeturn6view0turn20search0turn21view3turn21view0

This report treats four kinds of proposition separately. Formal findings are conclusions made under statutory authority by LECC, courts or other official bodies. Official observations and evaluations include descriptive analyses not amounting to misconduct findings or causal proof. Allegations and stakeholder accounts include claims by legal services, community organizations and journalists. Unresolved questions are matters for which no sufficiently detailed public data, policy or independent evaluation was located as of August 2, 2026.

Policy genealogy, target selection and operational design

The exact commencement date is described inconsistently in the public record. NSW parliamentary answers and LECC’s history place the initiative’s commencement in January 2000. BOCSAR’s evaluation describes the operational program as running since February 2002, apparently reflecting the beginning of the usable evaluation period or a later operationalized form. STMP-II was introduced in May 2005. These sources can be reconciled by treating 2000 as the policy’s institutional origin, 2002 as the beginning of the mature or data-observable operation described by BOCSAR, and May 2005 as the commencement of the second major version. citeturn0search5turn0search17turn2view2turn4search25

The intellectual genealogy lay in late-twentieth-century intelligence-led and “what works” policing. LECC reported that the initial design drew on the United States Congress-commissioned review Preventing Crime: What Works, What Doesn’t, What’s Promising. The operational premise was that a relatively small group of repeat offenders generated a disproportionate share of crime and that concentrating preventive and enforcement attention on them could produce deterrence or incapacitation more efficiently than undifferentiated patrol. The resulting model fused centralized policy with local selection: each Police Area Command or Police District could identify people linked to its priority crime problems and prepare individualized plans. citeturn12view0turn19view3

The secrecy surrounding the governing documents was itself a recurrent issue. NSW Police treated core policies, matrices and operational instructions as protected or classified information. Public knowledge came principally from parliamentary orders and questions, Government Information (Public Access) applications, litigation and disclosure to LECC under its compulsory powers. Operational sensitivity is legitimate where disclosure would reveal covert techniques or current intelligence, but the lack of public criteria prevented affected people, communities and Parliament from assessing the validity of the risk model, correcting errors or determining whether local commands applied consistent thresholds. citeturn0search5turn2view3turn16search7

STMP-I and STMP-II

The early plan’s object was to identify repeat or high-risk offenders and manage them with overt and covert attention. By the period examined in the Youth Justice Coalition and LECC interim reports, any police officer could initiate or suggest a nomination, although local intelligence staff performed the formal assessment. Local crime priorities mattered: the “crime environment” component asked officers to evaluate the importance of a suspected crime type to their command and the risk that the nominee would engage in it. The “person disposition” component considered multiple indicators intended to measure propensity to reoffend. citeturn24search4turn24search11

The STMP-II tool was described as automatically calculating a score and corresponding risk rating. “Automated,” however, referred to arithmetic performed after officers entered values; it did not mean that a machine independently discovered targets from the population. First, local police had already decided that the person might be suitable. Second, several inputs required interpretation of intelligence and suspected responsibility rather than adjudicated guilt. Third, local priorities influenced the crime-environment score. Fourth, approval and the operational response remained human decisions. The tool therefore standardized and quantified discretion without replacing it. citeturn24search4turn24search9

Publicly obtained STMP-II materials identified extreme, high, medium and low classifications. Classification was consequential because it helped structure the intensity and review of targeting, but it did not itself authorize any stop, search or entry. LECC found no evidence that the instrument had undergone rigorous scientific validation for either adults or children. It was especially concerned that an adult-oriented instrument compressed youth development into crude age indicators and did not capture the changing meaning of conduct across early, middle and late adolescence. citeturn24search4turn24search11

STMP-II plans could draw from four broad operational modes. “Positive” strategies included diversion, Police Citizens Youth Club activity or constructive engagement. “Compliance” included monitoring bail, parole or Youth Justice obligations. “Information gathering” involved intelligence collection and observation. “Overt” strategies involved visible intervention such as patrols around a home, stopping, searching, detaining or issuing consorting warnings. Local tasking and deployment meetings were expected to discuss targets at least fortnightly, and interactions were to be recorded in COPS. citeturn12view5

In October 2015, NSW Police created a domestic-violence version, commonly called DV-STMP, with implementation described in some official material as beginning in February 2016. It shared the idea of identifying repeat or high-risk persons but differed in its offense focus, victim context and notification practices. Because Operation Tepito’s principal concern was the general plan’s use on children, DV-STMP should not be treated as interchangeable with the youth program. citeturn4search25turn0search11turn23view2

STMP-III

STMP-III commenced on November 4, 2020 after the interim Operation Tepito report and internal reviews. It was presented as a more accountable, preventive and individualized framework. The old risk-assessment calculation was replaced by a Crime Severity Index and a mandatory comprehensive profile generated in Chimera, which drew together police data. An intelligence officer prepared a summary of offending history and a justification for selection; a senior local officer approved or rejected the nomination. The command then created a Target Action Plan specifying the objective, toolkit and approved strategies. citeturn0search23turn12view0turn24search4

The Crime Severity Index ranked recorded conduct by applying weights to police charges and, at lower weight, Events and Intelligence Reports. A court outcome did not retrospectively remove the weight attributed to a charge. This meant that dismissed, withdrawn or unproven allegations could continue to influence a score, although the index was only one part of the profile. LECC’s review of 128 available scores in its 133-person youth cohort found a mean of 1,177 and median of 953, but 38 young people had scores below 400, including five with zero, six below ten and twenty-one below 100. For eleven of those thirty-eight low-score cases, LECC considered the written nomination justification inconsistent with the underlying score or record. citeturn12view3

That distribution shows the distinction between computation and discretion. The index could rank the severity of recorded police contact, but policy did not make a minimum score an absolute gateway. Officers could regard intelligence, emerging conduct, local priorities or professional judgment as overriding. Flexibility can prevent blind reliance on a flawed model, but it also removes the discipline of an auditable threshold. LECC found that reasons for selecting people contrary to the index were not consistently clear and that the process insufficiently distinguished children from adults or tested whether less intrusive alternatives were preferable. citeturn12view2turn12view3

For young people, STMP-III required additional documents, including a Risk Factor Identification Toolkit. It was intended to capture matters such as family circumstances, education, health, disability, victimization and existing supports, and to inform whether a Youth, Prevention or Disruption approach was appropriate. In practice, police commonly completed the checklist from police-held information without interviewing the young person, family, carer or relevant service. LECC found known circumstances omitted from some profiles and little evidence that identified needs systematically changed selection or intervention. citeturn18view3

A separate Youth toolkit contemplated engagement, transport, referrals, Youth Action Meetings and coordination with youth-trained officers. The Prevention toolkit emphasized reducing opportunities or underlying contributors to offending. The Disruption toolkit emphasized overt enforcement and surveillance. The existence of three toolkits did not require a balanced mix; a Target Action Plan could be dominated by disruption, and officers outside the case-management chain could conduct interactions not listed in the approved plan if they independently believed ordinary powers were available. citeturn3view4turn11view6

The intended selection and intervention process can be reconstructed as follows:

Local crime priorities, COPS records, charges, Events, intelligence and officer observations
                                      │
                                      ▼
               Potential nominee identified by operational or intelligence police
                                      │
                                      ▼
       STMP-II: risk-tool inputs and automatic score/rating
       STMP-III: Crime Severity Index plus Chimera comprehensive profile
                                      │
                                      ▼
      Intelligence officer prepares offending history and written justification
                                      │
                        Youth-specific risk checklist
                                      │
             Additional senior approval for children under fourteen
                                      ▼
                 Senior local officer approves or rejects nomination
                                      │
                                      ▼
             Objective, toolkit and Target Action Plan are selected
              Youth / Prevention / Disruption strategies may include:
       engagement, referral, observation, intelligence, stop/search, home visit,
       bail check, consorting warning, FPO action, investigation or arrest
                                      │
                                      ▼
       STMP-III youth target should receive notification letter, ordinarily
                    documented through body-worn video
                                      │
                                      ▼
         General-duties and specialist officers conduct approved or other
       independently justified actions using ordinary statutory/common-law powers
                                      │
                                      ▼
          Encounters recorded in COPS; management documents held in Chimera
                                      │
                                      ▼
       Fortnightly local discussion and nominal three-month formal review
                                      │
                                      ▼
        Continue, modify, suspend during custody, transfer, or remove target

The diagram depicts policy rather than consistent practice. LECC found fragmented COPS and Chimera records, incomplete notification, actions absent from Target Action Plans, irregular case-officer review and inadequate articulation of legal grounds. citeturn12view0turn3view4turn19view3

Timeline and policy genealogy

Operation Tepito, police interactions and unequal impact

The cohorts

The interim Operation Tepito cohort consisted of all 429 people under eighteen placed on STMP over the two years from August 1, 2016. Police records marked 307—71.6 percent—as “possible Aboriginal or Torres Strait Islander.” LECC cautioned that the historical “possible ATSI” field was not a definitive self-identification measure, but the magnitude was nevertheless extraordinary. The cohort included children below the minimum age of criminal responsibility, and parliamentary evidence referred to a nine-year-old having been listed at an earlier point. citeturn2view1turn19view3

The final-report cohort comprised all 133 young people carried from STMP-II into STMP-III or added during STMP-III’s first six months. The average age was sixteen, and none was under fourteen, reflecting tighter approval rules. Sixty-four—48.1 percent—were recorded as Aboriginal or Torres Strait Islander, even though Aboriginal and Torres Strait Islander children represented approximately 4.5 percent of the NSW youth population. Four young people had no accessible STMP COPS case number, so some record-based analyses used 129 rather than 133 as the denominator. citeturn19view0turn3view1turn18view3

The fall from 71.6 percent in the interim cohort to 48.1 percent in the final cohort should not be interpreted as proof that racial disparity had been solved. The cohorts used different identification fields and periods; under-fourteen selection had become harder; and overall youth numbers had contracted. A group representing roughly one in twenty NSW children still accounted for almost one in two young STMP-III targets. LECC characterized the continuing disparity as gross overrepresentation and considered the selection process likely to have contributed to an improperly discriminatory effect. citeturn12view2turn3view6

Several pathways could generate that disparity without an explicit ethnicity variable in the score. Aboriginal children are already disproportionately exposed to police contact, charges, bail, public-order enforcement, socioeconomic disadvantage, out-of-home care and criminal-justice surveillance. A tool built from police records therefore imports prior enforcement patterns. Local professional judgment may then amplify the same pattern when police interpret intelligence, associations, family context and “visibility” in public space. Finally, STMP attention creates additional records, searches, technical-bail detections and intelligence, which may feed future selection. This is a cumulative-feedback mechanism, not proof that every individual decision was consciously race based. LECC’s finding was principally about discriminatory effect, structural bias risk and absence of adequate countermeasures. citeturn3view6turn10view5

Notification and procedural opacity

STMP-III’s youth process contemplated notification as the first targeting action. Police were to provide a letter explaining that the young person had been selected, describe relevant strategies and ordinarily capture the interaction on body-worn video. Notification potentially served both deterrence and fairness: a person cannot respond to a deterrent message, correct information or seek assistance if unaware of the status and expectations imposed on them. citeturn3view3

Implementation was poor. LECC found a notification letter or evidence of notification for only 54 of 133 young people, or 41 percent; 79, or 59 percent, had neither. Of the 54 letters located, 31 were not signed by the young person and 35 lacked a parent or guardian signature. Some were incomplete, and the strategies listed often did not correspond with what police actually did. A signature is not the only way to establish notice, but the combination of missing letters, incomplete records and inconsistent actions defeated both accountability and a coherent deterrence message. citeturn3view3

The program did not provide an external merits-review process through which a target could challenge selection, demand the evidence, correct a score or obtain an independent decision. Nor did inclusion create a public legal status comparable to bail, parole or a court order. Affected people therefore faced a paradox: the plan could shape police attention in highly consequential ways, yet its intelligence classification, reasons and operational details remained largely inaccessible.

There was no fixed statutory maximum duration. Policy expected formal review every three months and local discussion roughly fortnightly. A person could be removed after ceasing offending, transferred, or have active management suspended during custody. The public reports do not provide a complete distribution of time on STMP, numbers removed at each review, or reasons for cessation. The absence of an enforceable expiry period made the quality of internal review especially important. citeturn12view0turn11view5turn3view5

LECC found reviews and supporting records inconsistent. Some case officers periodically summarized contacts; others produced brief entries such as “STMP check,” “checked nil issues,” or “attended spoke to POI nil adverse.” Bail checks were frequently absent from the STMP case even where an Event appeared elsewhere in COPS. The division between Chimera, which held the plan, and COPS, which frontline officers used, meant that an officer encountering a target might not see the approved objective or strategies. This fragmentation undermined meaningful three-month assessment because reviewers could not reliably determine what had been done, why, with what result, and at what cost. citeturn19view3turn3view4

What “proactive policing” meant in practice

The policy’s public-facing description encompassed positive engagement, prevention, information gathering and enforcement. Its practical center of gravity was disruption. Of the 133 young people, 57—43 percent—experienced police actions that LECC considered inconsistent with the stated objective or approved Target Action Plan. Sixty-four—48 percent—experienced only Disruption-toolkit actions during the reviewed period. The most common encounters were stops and searches, home visits and bail-compliance checks. citeturn3view4

Police searched 84 of the 133 young people, or 63 percent, a total of 457 times. That is an average of 5.4 searches among those searched, although the distribution was heavily concentrated: the fifteen most frequently searched young people accounted for 274 searches, or 60 percent. For 59 of the 84 searched people—70 percent—neither the notification letter nor Target Action Plan identified stop and search as an intended strategy. Only 35 Target Action Plans listed search, and only five corresponding notification letters did so. citeturn11view1turn19view3

An action’s omission from the plan does not itself make it unlawful. A constable who develops reasonable suspicion during an encounter may use a statutory search power whether or not the case officer anticipated it. The concern is the opposite inference: being on STMP could not itself supply reasonable suspicion. LECC found records in which prior history, generalized intelligence, location or target status appeared to substitute for a contemporaneous factual basis. Searches frequently found nothing, and some records were too sparse to establish why suspicion was reasonable at the time. citeturn3view4turn15search0

Home visits were even more widespread. Police visited the homes of 104 young people, or 78 percent. Only fourteen Target Action Plans and four notification letters listed home visits, and just two young people had home visits included in both. Plans usually gave little direction about frequency, timing, purpose, consent or when to stop. Repetition thus depended substantially on local practice and frontline judgment. citeturn11view2

The intrusion extended beyond targets. Seven cohort members lived in out-of-home care, five of them Aboriginal. Police used bail checks for five and STMP home visits for six. One care residence received 65 STMP visits between April 28 and September 1, 2021, including 44 after 9 p.m., even though home visits were not an approved strategy for that target. LECC found no indication that officers assessed the effect on other traumatized children and staff or applied the government’s joint protocol favoring supportive, trauma-informed responses in residential care. citeturn11view0turn18view3

Bail checking was the most frequently recorded category of STMP interaction in a 2021 governance audit. Eighty-one young people—61 percent—were checked at least once, with 1,703 checks across the cohort. Fifty-five percent were absent from the corresponding STMP case record. One young person received 97 checks in less than twelve months; another received ten in fourteen days; another experienced eleven in seven days, including four within a single twenty-four-hour period. Seven young people experienced at least three checks in one day. citeturn19view3turn3view5

Thirty-two young people were arrested at least once for breach of bail while on STMP. Twenty-three of those thirty-two—72 percent—were arrested on one or more occasions for a “technical” breach rather than a new substantive offense. LECC identified 54 technical-breach arrests in total. One person was arrested ten times for technical breaches in twelve months; another, after 84 checks, was arrested seven times. Technical breaches can involve genuine risks and court-ordered obligations, but high-frequency checking mechanically increases the probability of detecting curfew, residence or reporting noncompliance that would otherwise remain unknown. citeturn11view3

This creates an important criminological distinction between offending incidence and police-detected events. If one group is visited, watched and checked many more often than otherwise comparable people, police will detect more possession offenses, public-order conduct and technical breaches even if the groups’ underlying behavior is identical. An increase in detected events after targeting cannot automatically be interpreted as increased offending. Conversely, a decline in selected offenses may coexist with a rise in technical enforcement because the program changes both behavior and observation.

Positive action was not wholly absent. Fifty-two young people—39 percent—had at least one positive strategy documented as applied. Fourteen—11 percent—were referred to a Youth Action Meeting or received an identifiable interagency referral. For 68—51 percent—police recorded that referral or engagement had been considered, but LECC found no evidence that it occurred; in some cases the young person did not consent. This gap between “considered,” “planned” and delivered assistance is central to assessing claims that STMP was a support program. citeturn19view2

Youth-trained officers appeared to use a different mix of actions. In LECC’s limited sample they averaged approximately two searches and three home visits per target, compared with about five searches and sixteen visits by general-duties officers, and they more often documented transport, referral and interagency work. The sample was too small and non-random to establish causation, but it supports the inference that training, role orientation and organizational setting influenced whether “proactivity” meant engagement or disruption. citeturn11view6

The public reports do not provide complete aggregate counts for every requested interaction. They quantify searches, home-visit exposure, bail checks, bail-breach arrests and referrals, but not all stop-and-speak encounters, general arrests, consorting warnings, cautions, move-on directions or removals from the program. Recorded interactions also understate actual activity where officers failed to link events to the STMP case. Any claim to a complete statewide encounter count would therefore be unsupported.

Children, complex needs and Indigenous justice

Youth offending is developmentally concentrated and often transitory. Impulsivity, peer influence, family instability, school disengagement, trauma, disability and victimization have different significance in adolescence than in adulthood. NSW’s Young Offenders Act is structured around warnings, cautions, conferences, proportionality and diversion from formal proceedings. For children aged ten to thirteen, the common-law presumption of doli incapax requires proof that the individual understood the conduct was seriously wrong rather than merely naughty or mischievous. citeturn12view4turn14search8turn15search15

STMP did not itself determine criminal responsibility, but targeting children below fourteen sat uneasily with those principles. A program could repeatedly search or visit a child because police predicted future offending even where the prosecution of particular alleged conduct would require proof of developmental capacity. Parliament’s concern about nine- and ten-year-old targets was therefore not simply symbolic: it exposed a mismatch between a preventive-risk model and the criminal law’s individualized protections for immature children. citeturn19view3turn15search11

LECC found many young targets had complex needs known to police, including disability, mental-health concerns, prior victimization, family dysfunction, substance use, educational disengagement and out-of-home care. The policy told case officers to consider detrimental effects and proportionality but provided little operational guidance on when those needs should make STMP unsuitable, how tactics should be adapted, or how consent-based services would be secured. citeturn18view3

Aboriginal Cultural Liaison Officer participation was particularly weak. Although policy and training contemplated ACLO input, only ten of the sixty-four Aboriginal young people—16 percent—had a Target Action Plan showing it. Only twenty-seven—38 percent—were managed under the Youth toolkit, while disruption was more common for Aboriginal than non-Aboriginal targets. Thirty-nine of sixty-four Aboriginal children—61 percent—experienced at least one Disruption action. citeturn3view6

For Indigenous communities, repeated police attendance has a historical and collective dimension. Stakeholders told LECC that highly visible targeting compounded mistrust produced by earlier overpolicing, removals and criminalization. That is a stakeholder observation rather than a finding that every contact caused harm. LECC nevertheless accepted that the program lacked practical strategies to address overrepresentation and that intensive surveillance could further marginalize young Aboriginal people. citeturn19view3turn7search10

A defensible future policy would therefore have to do more than remove ethnicity from a scoring formula. It would require Indigenous governance in design and review, community-controlled alternatives, measurement of every stage from nomination to coercive action, and authority to halt a program whose cumulative effects are disproportionate. Formal “race blindness” is inadequate when the input data and institutional environment are already stratified.

Effectiveness evidence and statistical appraisal

What BOCSAR evaluated

BOCSAR evaluated STMP-II and DV-STMP using linked police, court and custody records. For STMP-II it created matched comparison groups using coarsened exact matching and propensity-score matching, grouping placements by time and aligning comparison observations to treated start dates. Outcomes included whether a person recorded selected violent or property offenses in the following year, counts of those offenses, domestic-violence outcomes for DV-STMP and imprisonment. The full STMP-II cohort used for the custody analysis contained 10,122 people, including 2,746 juveniles and 4,617 Aboriginal people. citeturn13view0turn13view2

The published evaluation reported a large reduction in selected property offending associated with STMP-II, no comparable adult violent-crime reduction, and a rise in the probability of imprisonment. Associations were stronger for juveniles and weaker for Aboriginal people than for non-Aboriginal people. DV-STMP was associated with lower recorded domestic-violence offending without a statistically significant rise in imprisonment, although BOCSAR cautioned that victim reporting could be affected by notification or offender pressure. citeturn13view3turn23view2

A descriptive before-and-after graph showed roughly 9,000 selected violent or property offenses in the year before STMP-II placement and about 7,000 in the year afterward. BOCSAR divided participants by custody exposure and estimated that about 85 percent of the recorded reduction was attributable to cohorts spending no more than thirty days in custody, which it interpreted primarily as deterrence, and approximately 15 percent to the heavily incarcerated cohort, interpreted as incapacitation. citeturn23view0turn23view1

That interpretation is suggestive, not definitive. “Not in custody” does not isolate deterrence: people may mature, receive community supervision, move, enter treatment, experience reduced opportunity, or regress from an unusually high pre-selection period. Nor does custody duration arise randomly; it is affected by offending, bail enforcement, sentencing and police attention. The cohort decomposition cannot by itself identify the causal mechanism.

Quantitative results and limitations

The following confidence intervals are approximate 95 percent intervals calculated from BOCSAR’s reported coefficients and robust standard errors; they were not separately printed as intervals in the table.

The strongest threat to causal inference is selection bias combined with regression to the mean. STMP nominees were selected because recorded offending or intelligence indicated escalating risk. BOCSAR’s own graphs showed a steep rise immediately before placement, followed by a sharp fall. Even if STMP had no effect, selecting individuals near a temporary maximum would tend to produce lower subsequent values. BOCSAR argued that continued pre-placement growth made spontaneous decline implausible and might cause the study to underestimate benefit. Critics responded that a turning point is exactly what one expects when selection is triggered by unusual recent activity, and simulation work reportedly reproduced a similar pattern in synthetic data with no treatment effect. citeturn13view7turn1search8turn1search12

Matching improves balance on observed variables; it does not solve selection on unobserved intelligence, officer judgment, timing triggers or recent offense trajectories unless those are fully modeled. A person with the same annual offense count as a target but whose offending is stable or declining is not an adequate counterfactual for a nominee chosen because police perceived a sudden escalation. Quarter-based matching and fixed effects reduce some confounding but do not create random assignment. BOCSAR ultimately accepted this point by revising the publication to state expressly that the estimates lack causal interpretation. citeturn13view0turn23view3

A second problem is changing police attention. STMP increased the probability that minor offenses, possession, consorting, public-order conduct and technical bail breaches would be observed. That can produce an apparent increase in offending or criminal-justice contact even if behavior is unchanged. BOCSAR tried to reduce detection bias by emphasizing selected violent and property offenses thought less likely to be police-generated. That is a sensible design choice, but not a complete solution: vehicle theft, burglary or assault clearance can still be affected by intelligence focus, suspect attribution, surveillance and investigation. citeturn23view2turn8search10

The reverse problem also exists. A decline in police-recorded crime may not equal a decline in victimization. Targets may offend in other locations or ways, victims may report less, or police may classify events differently. BOCSAR expressly acknowledged this risk for DV-STMP, where notification might facilitate pressure on victims not to report. The evaluation did not measure victim surveys, community harm, school engagement, housing stability, wellbeing, trust, complaints or collateral effects on families. citeturn23view2

A third difficulty is incapacitation and intervention contamination. STMP-II was associated with increased imprisonment, including for juveniles and Aboriginal people after BOCSAR corrected its data error. A crime reduction obtained through more custody has different legal, fiscal and social implications from one obtained through voluntary desistance. Participants also received bail conditions, community orders, court proceedings, Youth Justice supervision and ordinary policing, so STMP was not a discrete treatment. citeturn23view3turn13view2

A fourth issue is outcome selection. BOCSAR’s principal result was stronger for property crime than violence. Describing STMP generally as “effective” obscures that heterogeneity. A program may reduce a selected recorded outcome while increasing imprisonment, technical breaches, police contacts, trauma or offending in a subgroup. The positive coefficient for an Aboriginal custody subgroup does not prove that STMP caused increased offending, but it warns against assuming uniform benefits. citeturn23view1turn23view2

A fifth concern is implementation variation. BOCSAR estimated an average association across many commands and years, while Operation Tepito showed wide variation in notification, contact intensity, youth expertise, recordkeeping and balance between support and disruption. “STMP” was not a single standardized dose. One target might receive referral and occasional engagement; another might receive dozens of visits and checks. Without reliable intervention-intensity measures, an average effect cannot identify which components helped, harmed or did nothing.

The fair conclusion is that the BOCSAR study constitutes evidence of a robust temporal association, not proof that STMP caused the reported decline. Its size and administrative coverage make it more informative than anecdote, while its non-random selection and sharp pre-placement trajectories prevent a confident deterrence estimate. The Operation Tepito evidence, meanwhile, is stronger on what police actually did and whether governance worked, but was not designed to estimate crime reduction. The two bodies of evidence answer different questions and should not be used to cancel each other out.

Police powers, public law, rights and remedies

STMP was an internal administrative policy, not a source of coercive authority. This was stated in the guidelines and repeatedly emphasized by LECC. The legal question in each encounter was not “Is this person an STMP target?” but “What power, consent or common-law permission authorizes this particular action now?” citeturn10view2turn19view3

Stops, searches and identity demands

Section 21 of the Law Enforcement (Powers and Responsibilities) Act 2002 permits a warrantless search of a person where an officer suspects on reasonable grounds that the person possesses stolen or unlawfully obtained property, something used or intended for an offense, a dangerous article in a public place, or a prohibited plant or drug. Comparable provisions govern vehicles and other contexts. LEPRA also regulates identity demands, move-on directions, arrest, use of force, strip searches and safeguards requiring officers to provide reasons and identification in specified circumstances. citeturn14search0turn14search12

In Rondo v R, the NSW Court of Criminal Appeal explained that reasonable suspicion is less than belief but more than a mere possibility and requires some factual basis; the source and content of information must be assessed for reliability. A person’s criminal record can contribute to suspicion, but status as a known offender or STMP target cannot alone establish possession of a prohibited item at a particular time. citeturn15search0turn15search32

This is where STMP’s design created systemic legal risk. The plan directed officers to seek interactions with a named person. Once an officer approached expecting to search, confirmation bias could transform neutral behavior, location or prior history into purported reasonable suspicion. The legal test remained event specific, but the operational instruction was person specific. LECC found examples in which records did not demonstrate a sufficient contemporaneous basis and warned that policy status had become confused with legal authority. citeturn18view3turn3view4

Move-on and arrest powers present similar risks. DPP v Carr illustrates that arrest for a minor offense may be improper where a summons or less intrusive process would suffice, with consequences for evidence under section 138 of the Evidence Act. STMP’s disruption objective could create an organizational incentive to arrest or issue directions even where necessity and proportionality pointed elsewhere. citeturn15search1turn15search5turn14search13

Strip searches of children are governed by additional LEPRA protections and must satisfy strict necessity and seriousness requirements. Operation Tepito’s principal aggregate findings concern ordinary searches rather than a quantified STMP strip-search cohort, so it would be wrong to infer that all or most of the 457 searches were strip searches. The governing principle remains that an intelligence plan cannot lower the statutory threshold.

Home visits and entry

No provision authorized police to enter a home merely because an occupant was on STMP. Entry required a warrant, a specific statutory power, informed consent, an emergency doctrine or a common-law permission. Police, like other members of the public, ordinarily have an implied licence to take the usual route to a front door and knock for lawful communication. In Roy v O’Neill, a High Court majority held that police could rely on that licence where their purpose included lawful communication and was not solely to impose coercive process. The licence is limited, can be withdrawn, and does not itself authorize a search or compelled entry. citeturn15search6turn15search2

Repeated late-night visits test the boundaries of that principle. A social licence to knock once for ordinary communication is not necessarily an unlimited authority to return dozens of times, ignore withdrawal of consent or use attendance as coercive surveillance. LECC found cases in which families withdrew consent and COPS warnings were added, yet conduct and documentation remained problematic. Whether a particular visit constituted trespass would depend on facts and ultimately be determined by a court. citeturn19view0turn10view2

Bail checks

The Bail Act 2013 permits courts to impose enforcement conditions authorizing specified police checks where necessary and reasonable. Section 81 permits directions under an enforcement condition where an officer reasonably suspects noncompliance. The statutory scheme allows courts to determine the kind, timing and frequency of enforcement after considering risk and proportionality. citeturn14search1turn14search4

Operation Tepito found that STMP practice frequently treated bail as a general opportunity for home attendance, sometimes without clarity about whether a court had imposed an enforcement condition or whether police relied instead on implied licence. LECC’s separate 2025 bail-compliance report concluded that NSW Police should seek properly framed section 30 enforcement conditions where entry onto private property is needed and should discontinue reliance on implied licence for such checks, or alternatively seek legislative amendment. It described the argument that the Bail Act displaces the implied licence as powerful while acknowledging that definitive resolution belongs to the courts. citeturn24search12turn7search3

NSW Police’s 2025 public account of its standard operating procedures continued to describe checks based on the Bail Act and implied licence: officers were to activate body-worn video, take the direct route to the front door, knock, confirm the person and compliance, then leave, subject to court-imposed limits. This indicates that legal disagreement persisted after STMP’s abolition. citeturn21view3

Questioning and interactions with minors

A child who is under arrest or otherwise detained under Part 9 of LEPRA receives rights concerning communication with a lawyer, friend, relative, guardian or support person, subject to statutory qualifications. Section 13 of the Children (Criminal Proceedings) Act generally protects admissions by children under sixteen unless made in the presence of an appropriate adult or otherwise within the statutory exceptions. The Evidence Act also permits exclusion of admissions for unfairness and of improperly or illegally obtained evidence where the desirability of admission does not outweigh the undesirability of the method. citeturn14search2turn14search3turn14search13

Not every doorstep conversation or street encounter amounts to detention or formal questioning. That creates practical ambiguity for young targets: a child may perceive a uniformed request as compulsory even where police regard the interaction as voluntary. Repeated contacts can also generate intelligence without the safeguards attached to arrest. A future framework should require officers to state clearly whether participation is voluntary, avoid intelligence questioning without a support person for children, and record the basis and purpose of every targeted encounter.

Administrative law and procedural fairness

STMP inclusion was not a court order and apparently did not directly alter a person’s legal rights or duties. That makes conventional judicial review difficult. An applicant would need to identify a reviewable administrative decision, legal error, denial of procedural fairness, improper purpose, unreasonableness or unlawfulness in implementation. Because the plan guided operational discretion and relied on protected intelligence, a court might be reluctant to supervise its merits. No reported judgment located for this report directly invalidated an STMP nomination or declared the statewide policy unlawful.

The absence of direct legal effect does not mean inclusion was inconsequential. It predictably exposed people to concentrated surveillance and increased opportunities for coercive encounters. Contemporary administrative-law principles can recognize practical interests beyond formal rights, but whether procedural fairness is implied depends on statutory context and the nature of the power. STMP’s non-statutory character, secrecy and asserted preventive purpose left this question unresolved rather than affirmatively answered.

At minimum, public-law legality required NSW Police to follow the limits of its governing statutes, consider relevant matters, avoid inflexible application, and not use an internal plan for an unauthorized purpose. Departures from internal policy do not automatically invalidate police conduct, but systematic disregard of plans, review requirements and youth safeguards supported LECC’s agency-maladministration analysis. citeturn16search2turn3view7

Procedural fairness was particularly weak in four respects: many targets were not effectively notified; reasons and underlying records were not disclosed; no independent merits review existed; and there was no fixed maximum duration. Intelligence confidentiality can justify redaction, but it does not require complete absence of a contestable summary, correction procedure or periodic external review.

Privacy and information law

STMP processed highly sensitive personal information: allegations, charges, intelligence, family relationships, addresses, disability, mental health, victimization, associations and predictions about future conduct. Under section 27 of the Privacy and Personal Information Protection Act 1998, NSW Police is broadly exempt from the Act’s information-protection principles except in relation to administrative and educative functions. That significantly limits the Information and Privacy Commission’s capacity to provide ordinary privacy-law remedies for operational intelligence practices. citeturn16search11

GIPA offers a presumptive right of access to government information, but law-enforcement, public-safety, confidential-source, operational and personal-privacy considerations can establish an overriding public interest against disclosure. GIPA litigation and parliamentary production were nevertheless important in exposing historical policy material. A future risk-management regime should not rely on these general access mechanisms as its primary transparency safeguard; it should require proactive publication of criteria, data dictionaries, validation, audit outcomes and aggregate demographic effects. citeturn16search7turn2view3

The use of charges and intelligence without regard to final court outcome also raises data-quality concerns. An arrest or charge is not a conviction, and intelligence may be incomplete or wrong. A defensible system needs correction, provenance and expiration rules; it should distinguish allegations, findings of guilt, victim records, association data and officer opinion rather than collapsing them into a single risk narrative.

Equality, children’s rights and Indigenous rights

The Anti-Discrimination Act 1977 prohibits race discrimination in specified fields, including the provision of goods and services, while the federal Racial Discrimination Act 1975 contains broader protections against race-based impairment of rights. Applying those statutes to an individual police targeting decision would require proof of the relevant conduct, field, causation or indirect-discrimination elements, and any applicable exceptions. Systemic disparity alone does not establish liability in a particular case, but it is powerful evidence requiring explanation and mitigation. No reported STMP merits judgment was located that resolved these questions. citeturn16search1

Australia’s obligations under the Convention on the Rights of the Child include treating the child’s best interests as a primary consideration, hearing children in matters affecting them, protecting privacy and administering youth justice consistently with dignity, reintegration and the child’s constructive role in society. The International Covenant on Civil and Political Rights protects against arbitrary interference with privacy and arbitrary detention and guarantees equal protection. The Convention on the Elimination of Racial Discrimination addresses racially discriminatory effects. The United Nations Declaration on the Rights of Indigenous Peoples adds standards concerning equality, culture, participation and special protection for Indigenous children. These instruments are not, without incorporation, free-standing causes of action invalidating NSW police conduct, but they inform policy, statutory interpretation and Australia’s international accountability.

Measured against those norms, STMP’s weaknesses were not merely technical. Missing notice limited participation; repeated home attendance interfered with privacy; targeting children without a validated youth model conflicted with developmental treatment; and Aboriginal overrepresentation without shared Indigenous governance undermined equality and self-determination. The state’s Young Offenders Act and Aboriginal Strategic Direction already pointed toward diversion, least-restrictive responses and improved relationships, yet LECC found the STMP documents and practices insufficiently integrated with them. citeturn12view4turn3view6

Remedies and oversight

A person subjected to an unlawful search, arrest, detention or entry may seek exclusion of evidence under section 138 of the Evidence Act and, depending on the facts, civil remedies for false imprisonment, assault, battery, trespass or malicious prosecution. Courts can issue declarations, injunctions or judicial-review relief where a reviewable legal error is established. Damages litigation is fact intensive and generally addresses an individual incident, not the validity of a confidential statewide classification. citeturn14search13turn15search5

Complaints could be made to NSW Police or LECC. LECC can investigate serious misconduct and systemic maladministration, compel information and report to Parliament, but it does not ordinarily determine civil compensation. Anti-Discrimination NSW and the Australian Human Rights Commission provide complaint avenues within their statutory jurisdictions. GIPA and privacy review offer limited information remedies. Parliament can question ministers, conduct committee inquiries and order document production. None supplied a simple, timely appeal against placement.

The government did not establish a dedicated STMP redress or compensation scheme. The Premier reportedly ruled out a general compensation response after the final report, while parliamentary answers referred affected people to existing complaint and legal processes. That leaves remedies dependent on proving particular unlawful encounters, often years after events and despite deficient records. citeturn5search0turn7news20

Case-law and oversight chronology

Australian analogues, place forecasting and international comparisons

Place-based forecasting

Person-based targeting should be distinguished from place-based forecasting. Australian Institute of Criminology researchers have tested whether historical crime data can forecast short-term burglary, motor-vehicle theft and theft-from-vehicle locations in Queensland. Using nine years of data across three local-government areas, the research found that several algorithms could achieve comparable or better “hit rates” with less historical data, but emphasized that optimal parameters varied by place and offense. The work evaluated forecasting performance, not whether sending police to predicted locations reduced crime without displacement or unequal enforcement. citeturn9search0turn9search1

Place forecasting has a different unit of analysis and intervention. It predicts where an event may occur, not which named individual will offend. Yet it can generate similar feedback loops: recorded crime sends patrols to an area; increased patrol produces more detected crime; the new records reinforce the forecast. The AIC research specifically recognized data-quality and feedback concerns and the need for long-term, locally validated evaluation before broad implementation. citeturn10view5

AIC work has also examined machine-learning identification of high-harm offending among outlaw motorcycle-gang cohorts. That research is closer to person-based prediction but remains an analytical study rather than proof of an operational national target program. It emphasizes the dependence of prediction on police-record quality and the limited portability of models across populations. citeturn9search4turn9search10

Other Australian jurisdictions

Victoria Police has used youth and offender-management practices including a Youth Networked Offender database, Priority Target Management Plans and Operation Wayward, which focused on young people associated with aggravated burglary and home invasion. Public descriptions refer to monitoring, case management and coordinated responses for people approximately ten to twenty-four years old. Publicly accessible statewide demographic and causal evaluations are limited, making it impossible to determine whether the scale, scoring or interaction intensity matched NSW STMP. citeturn12view1

Queensland introduced a Serious Repeat Offender Index and associated multi-agency management for high-frequency young offenders. Its model is more explicitly connected with youth-justice and service agencies, although Queensland’s later legislative “serious repeat offender” framework and punitive youth reforms complicate comparison. An index can improve consistency but also reproduce police and court-contact disparities unless selection, overrides and outcomes are independently audited. citeturn12view1

Victoria’s Scan–Prioritise–Manage and offender-management projects have employed focused-deterrence concepts, while Australian jurisdictions also use specialized domestic-violence risk tools. These initiatives demonstrate that risk assessment is not unique to NSW, but tools built for victim safety, place deployment or case coordination should not be equated with a person-level proactive-policing list. citeturn9search18turn9search24

No publicly available evidence located for this report establishes that another state or territory operated, as of August 2026, a direct statewide equivalent with STMP’s combination of individual risk rating, local target plan, routine stop/search attention and long-term statewide use.

Facial recognition and real-time intelligence

Facial recognition answers an identity question: whether an image matches a person in a database or watchlist. STMP answered a risk-management question: whether police believed a person likely to reoffend and what attention should follow. Facial recognition may help locate an STMP-like target, but it does not by itself predict offending. Its distinctive risks include false matches, demographic accuracy, watchlist legality, mass biometric collection and real-time tracking.

Credible reporting in July 2026 described a Western Australia Police real-time facial-recognition trial using a watchlist of roughly 4,000 people including wanted persons, registered child-sex offenders and missing people, and reported an arrest generated during the launch period. Those reports concern identification against a defined watchlist, not a disclosed algorithm predicting future offending. The distinction matters because the legal justification for including a person on a watchlist and the consequences of a match require separate scrutiny. citeturn9search17

A real-time crime center is an organizational and technological hub integrating dispatch, CCTV, automatic number-plate recognition, intelligence databases, maps and live incident information. It can support reactive investigation, place deployment or person targeting. NSW Police’s State Intelligence Command publicly describes 24-hour operational support for serious crimes and real-time incidents, analytical support and information sharing. No public evidence reviewed here shows that an American-style real-time crime center automatically generated STMP targets or that PDR currently operates a comparable person-risk algorithm. citeturn20search1turn20search13

Chicago, Los Angeles and Pasco County

These comparisons produce several lessons. First, person-based programs routinely blur victim, suspect and future-offender categories. Chicago’s list predicted involvement in violence, not guilt, yet operational users could treat a high score as suspicion. Second, human discretion does not eliminate algorithmic bias; LASER and STMP show that officer nominations, intelligence reports and overrides can be less transparent than a fixed formula. Third, intensive police attention changes measured outcomes, especially arrests and technical violations. Fourth, systems often persist for years without credible counterfactual evaluations because “proactivity” is treated as self-validating. citeturn8search0turn8search1turn8search10

STMP differed from all three in important ways. It ran statewide and for much longer; it used ordinary NSW powers rather than United States stop-and-frisk or probation doctrines; and its later version gave human decision-makers more control than Chicago’s ranked list. Yet the recurring failure mode was the same: a risk designation intended to allocate attention became an operational reason for attention, and attention produced the records used to validate continuing risk.

Discontinuation, successor practices, safeguards and final judgment

Why STMP ended

The termination was not the result of a single court ruling or a definitive randomized evaluation. It followed cumulative scrutiny: Indigenous and youth legal-service concerns from at least 2015; the 2017 Youth Justice Coalition report; parliamentary questioning; internal police reviews; LECC’s 2020 interim recommendations; the BOCSAR evaluation and methodological dispute; and LECC’s detailed STMP-III record review. citeturn7search4turn23view3turn18view3

NSW Police initially responded through reform rather than abolition. It raised approval for young children, created STMP-III, introduced Chimera profiles, changed notification material, established an internal review committee, reduced youth numbers and proposed a further youth model from July 2023. Police told LECC that the reforms would improve assessment of complex needs and supported referral pathways. citeturn12view6turn18view3

LECC concluded that the central problems persisted. Target selection retained unacceptable bias risks; no validated youth prediction instrument had emerged; police often ignored complex needs; disruption overwhelmed positive intervention; legal powers were confused with policy direction; records were inadequate; and reviews could not rigorously test whether intervention remained useful. It recommended that NSW Police cease applying STMP to children. citeturn3view7turn19view3

By June 2023, forty young people remained on STMP. In September, the Commissioner advised LECC that the number had fallen to one and that the executive had decided to reconsider the entire approach. On October 3, NSW Police confirmed that STMP was no longer applied to young people and would cease for adults by December. The force committed to developing a contemporary, fit-for-purpose approach to serious crime and repeat offending and to engaging LECC. citeturn18view2turn18view3turn24search2

LECC chose not to make a formal agency-maladministration finding after cessation, while recording its opinion that the threshold was met for the reviewed youth period. This reflected remedial pragmatism, not exoneration. It also means that statements describing STMP categorically as “declared unlawful” overstate the legal record. Particular actions were found or suspected to be unlawful; the systemic application was assessed as maladministrative; the policy was withdrawn before a formal finding or direct merits judgment. citeturn1search5turn3view7

Prevention, Disruption, Response

The current public successor is PDR, implemented from around July 1, 2024 through standard operating procedures. LECC describes it as the framework through which local commands identify, respond to and monitor crime issues, including youth engaged in or at risk of offending, with activities linked to the Youth Strategy and Aboriginal Strategic Direction. NSW Police has shared some information with LECC and committed to involve it in review. citeturn6view0

The Police Minister initially described PDR as more local, grassroots, collaborative and connected with community involvement. In later parliamentary answers, NSW Police drew a sharper distinction: PDR “is not a program,” its procedures focus on strategic responses to crime issues, and individuals are not policed pursuant to it. Commands may consult community members, use engagement programs or invite community representatives to PDR meetings. citeturn21view2turn21view3

This distinction appears to reflect a shift from a durable person-centric list to an issue-centric command process. PDR meetings examine crime trends and develop prevention, disruption and response strategies. In February 2026, Deputy Commissioner David Hudson described a program or capability that examines crime data so commands can deploy taskings, with taskings adjusted daily according to trends, but rejected the proposition that NSW Police was mapping future crime through artificial intelligence to the degree suggested by the question. citeturn21view0

The distinction cannot yet be independently verified in detail because the operational procedures and internal analytics are not public. Parliament sought demographic counts of people “policed pursuant to PDR.” Police responded that no such population exists for reporting purposes and that ordinary individuals coming to attention are recorded in COPS, with no demographic data specific to PDR. This may accurately reflect an issue-based model. It also creates an audit problem: if interventions developed at PDR meetings are not tagged to PDR, neither Parliament nor LECC can readily test whether particular groups receive disproportionate disruption. citeturn20search0turn21view3

The internal six-month compliance review was completed but NSW Police said its results were not intended for external release. Police committed to involving LECC in a twelve-month review, and LECC’s 2024–25 annual report said it would continue monitoring proactive operations. As of August 2, 2026, no public independent effectiveness evaluation, demographic impact report or complete PDR standard operating procedure had been located. citeturn21view3turn6view0

PDR should therefore be described as the organizational successor, not necessarily a renamed STMP. There is no public evidence that it maintains the old extreme-to-low categories, Crime Severity Index, individual STMP status, notification letters or three-month person reviews. At the same time, prevention and disruption tasking can still generate repeated attention toward named people. The absence of a formal target list reduces some risks but can also make cumulative treatment harder to identify.

Model safeguards

Any future person-focused offender-management program should begin with legislation or a publicly issued statutory code rather than a protected internal policy. The purpose, eligible offenses, minimum seriousness, permissible data, decision-makers, duration, interventions and review rights should be specified. The statute should state expressly that inclusion creates no search, entry, questioning, arrest or bail-check power and cannot by itself constitute reasonable suspicion.

A defensible framework would incorporate the following safeguards:

These safeguards are not obstacles to effective policing. They are methods for distinguishing a genuinely evidence-based intervention from an unreviewable accumulation of police attention. A program unable to identify who is subject to it, measure its actions, justify disparities or test a counterfactual cannot reliably claim effectiveness.

Final judgment

Legally, STMP was not punishment in the formal sense. Inclusion was not a conviction, sentence, bail order or statutory disability. It did not itself authorize detention, search or entry. Properly used, it could operate as an internal resource-allocation and investigation framework, with each coercive action independently justified under ordinary law.

Functionally, however, surveillance and preventive control were its dominant characteristics. Named people were observed, stopped, searched, visited and checked because police assessed them as likely to offend in the future. The final Operation Tepito cohort experienced 457 searches, 1,703 bail checks and home visits for 78 percent of targets, while only 11 percent had an identifiable Youth Action Meeting or interagency referral. Nearly half experienced only disruption. citeturn11view1turn19view2turn19view3

Investigatively, STMP generated intelligence and could assist in solving or preventing particular crimes. Yet investigation normally begins with evidence of an offense; STMP often began with a prediction about a person. That orientation increased the risk that ordinary behavior would be viewed through a presumption of criminality and that past allegations would become a standing basis for future scrutiny.

Supportively, the program contained genuine positive options, and some officers used them. The evidence does not justify saying that every plan was punitive or that no young person benefited. But support was structurally secondary: referrals were frequently merely considered, required consent and resources not guaranteed by the plan, and were dwarfed by enforcement activity.

Sociologically and experientially, the most intensive applications resembled extra-judicial punishment. Ninety-seven bail checks in less than a year, sixty-five visits to a care residence, repeated searches finding nothing and multiple arrests for technical breaches imposed burdens on liberty, privacy, sleep, family life and reputation. Those burdens followed an administrative prediction rather than a judicial finding that additional punishment was deserved. Calling them preventive did not remove their punitive effect. citeturn3view5turn11view0turn11view3

The fairest final characterization is therefore:

STMP was principally a surveillance-and-disruption system, secondarily an intelligence and offender-management system, intermittently a support mechanism, and—at its most intensive—functionally punitive, even though it was not punishment as a matter of formal sentencing law.

Its central public-law failure was not simply that prediction can be wrong. It was that a fallible and partly discretionary prediction was converted into repeated real-world interventions without adequate notice, validated youth criteria, reliable records, Indigenous governance, independent review or consistent proof of legal authority. Its central criminological failure was treating increased police activity as an intervention whose benefit could be assumed rather than isolated. Its central statistical failure was the absence of a credible causal design before statewide, long-term implementation. Its central Indigenous-justice failure was allowing historical police-contact data and local discretion to produce extraordinary Aboriginal overrepresentation without effective shared control.

The lesson for PDR and any later successor is not that police must ignore patterns of repeat serious harm. It is that future risk cannot lawfully or scientifically function as a free-standing warrant for present coercion. Prediction may inform where police look and what assistance they offer; it must never silently dilute the legal thresholds governing what police may do.