Colombia's military false-positives system

The reported combat result began with a false record about a person

The phrase “false positive” compresses a killing and its administrative reversal: a civilian is killed unlawfully, the scene is staged to resemble combat, and the person is reported as a guerrilla or criminal killed lawfully. The United Nations Special Rapporteur's 2009 mission noted examples dating to the 1980s and a disturbing nationwide frequency from 2004, then described recurring variants: false job offers, paid recruiters or informers, removal from a home or roadblock, transport, killing, weapons or clothing placed on the body, press and military reporting, burial without identification, family searches, and threats against people seeking justice. The mission called these patterns general, not a claim that every case contained every step.¹

The JEP's figure of 6,402 is not a count of 6,402 adjudicated homicides. Auto 033 reported 6,402 unique victim records for 2002–2008 after matching and deduplicating four databases supplied by the ordinary prosecutor, the National Centre for Historical Memory, and civil-society organizations. The order says the number was not definitive, remained under contrast, and would be updated with testimony and military-justice data. It also adopted a bottom-up investigation from regional and local participants toward possible responsibility at other territorial and national levels. The JEP's public Case 03 profile continued to call the figure a provisional universe when accessed on 14 July 2026.²

That distinction preserves four things the euphemism can hide: a person rather than a score, an estimate rather than a verdict, an alleged or proven act rather than a job title, and individual legal responsibility rather than an inference from a national total.

Incentives made a reported death valuable without proving one universal order

The 2010 UN mission rejected two opposing accounts. It had not seen evidence that the killings followed an official policy or orders from senior government officials, but it found too many similar, credible cases across departments and units to accept an explanation of isolated soldiers or “bad apples.” It separately documented pressure to show results, civilian informer payments and discretionary funds, and reported informal rewards such as leave, medals, and promotion, while recording the Colombian government's disputes and policy explanations.³

A peer-reviewed municipality-level study found significantly more recorded false positives during its 2003–2008 high-powered-incentive period where local judicial institutions were weaker and a higher share of brigades were led by colonels. In those municipalities the period also coincided with worse local judicial quality and no discernible security improvement. The authors explicitly say their estimates are not causal effects and cannot exclude time-varying alternatives; the panel, CINEP-derived outcome data, no-pretrend tests, and robustness checks support an association, not proof of a top-down order or any person's liability.⁴

The organizational point is therefore narrower than “a metric caused murder.” People exercised criminal agency. Results pressure, career concerns, rewards, and weak external checking altered what a reported death could do for a unit or member; they do not erase intent, refusal, variation among units, or the legal elements that must be proved for each defendant.³⁴

Segmented work made a killing administratively repeatable

The Truth Commission reconstructed four overlapping, not necessarily linear, phases: planning, execution, concealment, and impunity. Its synthesis describes victim selection or recruitment, transport, killing, a “legalization kit” of weapons or other items, altered scenes, false operational, patrol, intelligence, forensic, ammunition, and reward records, identity removal, no-name burial, approval, and benefits. Roles could be split among recruiters, informers, transporters, soldiers, officers, clerks, forensic actors, military judges, and paramilitary participants; the composition differed across places and cases.⁵

This is evidence of coordinated work, not guilt by occupation. A document signer might have been deceived, complicit, resistant, or responsible; a recruiter or informer might appear in an allegation, testimony, or adjudicated event. The criminal inference must follow the evidence for that person's knowledge, conduct, authority, and case. Segmentation matters because it let participants encounter only part of the process while a complete combat account moved upward.⁵

The Commission's case study is a broad official historical synthesis. It processed 1,042 victim interviews, 178 interviews with JEP participants, 412 interviews with experts, witnesses, and other actors, 320 reports, and 91 submitted cases. Those inputs widen the record but are not one evidentiary category: testimony, an organization's report, an official database, and a judgment carry different claims and limits.⁶

Upward reporting concentrated success while contradiction stayed dispersed

A unit could produce the operation order, tactical narrative, weapons account, patrol report, photographs, and reward paperwork that made its own result legible. Contradictory evidence might instead sit with a family, missing-person office, morgue, cemetery, ordinary prosecutor, military court, or witness. UN follow-up records identify jurisdictional conflict, threats, pressure on military judges and cooperating service members, insufficient independent scene examination, resource limits, and the absence at that time of a central database showing the progress of criminal and disciplinary cases.⁷

Human Rights Watch's 2015 investigation used criminal files, judicial rulings, prosecution data, witness material, and more than forty interviews with prosecutors, witnesses, family members, and lawyers to argue that recurring methods and warning signs supported investigation of senior command responsibility. It also documented alleged obstruction and retaliation. Those were advocacy and investigative conclusions about then-current allegations and proceedings, not judgments against the officers discussed; later outcomes must be checked person by person rather than inherited from the report.⁸

Families were not merely inputs to an official control system. In a 2019 JEP hearing, thirteen Soacha relatives presented observations on military participants' accounts; the JEP record says MAFAPO had earlier become the first victims' organization to submit a report directly to Case 03. The Truth Commission identifies that September 2018 MAFAPO report among its direct inputs and describes women and families using their own emotional, physical, and financial resources to search, organize, challenge stigma, and assemble records across offices. The public sources located here mediate that family-controlled report through JEP and Commission records; they do not substitute for a stable MAFAPO-controlled archive or make families a single voice.⁹

Estimates, historical findings, accusations, and judgments answer different questions

The Truth Commission later reached a broader conclusion than the 2010 UN mission. It described a policy that was not formally adopted, concluded that the practice was systematic and generalized in 2002–2008, and assigned ethical, political, and historical responsibility under what it expressly called its extrajudicial mandate. That is an official truth-commission finding grounded in a later and larger record; it is not a criminal conviction, and it does not retroactively turn the UN mission's earlier statement that it had seen no evidence of an official policy into a judicial finding either way.¹⁰³

The JEP's 18 September 2025 La Popa acknowledgment-route sentence is narrower and judicial. It concerns 135 people in 77 events from January 2002 through July 2005. Of fifteen people called to acknowledge responsibility, twelve did so; the first-instance section determined those twelve to be among the most responsible in the subcase and imposed individualized restorative sanctions ranging from five years and seven months to eight years, with effective restrictions of rights and liberty. The operative order expressly permits appeal, so the cited judgment alone does not establish later finality.¹¹

The JEP's 17 December 2025 adversarial judgment is also defendant- and subcase- specific. A first-instance section found Publio Hernán Mejía Gutiérrez responsible beyond reasonable doubt for protected-person homicide, torture, and forced disappearance as war crimes and crimes against humanity within a systematic macrocriminal pattern. It treated thirty-five charged events involving seventy-two victims and imposed twenty years of imprisonment. The orders say capture may occur only when the judgment becomes final and expressly allow appeal. This is an adjudicated first-instance finding, not proof against every commander or a final national command-responsibility judgment.¹²

The evidence ladder is therefore explicit. A database estimate describes a provisional universe. A family or witness account supplies testimony and counter-record. An investigation or accusation remains allegation until adjudicated. A truth commission can make historical and political findings within its mandate. A first-instance court can establish criminal responsibility for a defendant and event set while leaving appeal open. None of those records licenses a shortcut from 6,402 estimated victims to 6,402 convictions, from similar method to one universal order, or from command position to individual guilt.

Reform reduced reported cases without closing accountability

After the 2008 Soacha revelations, the UN mission recorded dismissals, independent scene-investigation requirements, operational legal advisers, changes to award and unit-performance criteria, oversight of informer payments, specialized prosecution, and stronger transfer rules toward civilian justice. Allegations dropped sharply, but the 2010 report said it was too early to confirm the nature or extent of the decline. The 2012 follow-up accepted a drastic decrease while continuing to identify new cases, jurisdictional risks, opaque incentive monitoring, threats, forensic limits, and persistent impunity.¹³

The decline is evidence that institutional choices mattered, not proof that every measure worked as intended or that earlier cases were resolved. Prevention requires independent examination of every reported combat death, joined operational, missing-person, burial, and case-progress data with privacy controls, protected refusal and cooperation, victim participation, and individual adjudication at the appropriate evidentiary standard.

Comparisons identify Atlas mechanisms, not equivalence or ancestry

Measurement, accounting, and control is an idea relation: reported combat deaths became a performance signal, and the same units could help generate the record that validated the signal. Governance, stewardship, and accountability is an idea relation about independent review, jurisdiction, command responsibility, remedy, and whose account can correct an official result.

Culture, informal organization, trust, and voice is an idea relation about pressure, rewards, stigma, solidarity, resistance, and the safety of contradiction. Executive attention, information, and organizational sensing is an idea relation about which results moved upward, which warning signs leaders could inspect, and how distributed family and judicial evidence reached decision makers. These are Atlas analytical relations, not historical vocabulary, documented influence, equivalence to another case, or a claim that one mechanism alone explains criminal responsibility.¹⁴

The comparative fingerprint makes emphasis inspectable

Score 3 marks purpose, mission, and institutional legitimacy, authority, legitimacy, and acceptance, coordination, communication, and common understanding, measurement, accounting, and control, cooperation, incentives, and organizational equilibrium, and governance, stewardship, and accountability. The criminal system inverted the protective mission, used military and bureaucratic authority, coordinated segmented work, made deaths countable, distributed rewards and pressure, and disabled or delayed correction.¹⁴

Score 2 marks delegation, decentralization, and responsibility, structure, hierarchy, and scale, decision-making, judgment, and bounded rationality, learning, quality, and reliability, culture, informal organization, trust, and voice, executive attention, information, and organizational sensing, and organizational ignorance. Local variation, ranked hierarchy, repeated judgment, learned concealment, informal pressure, vertical reporting, and advantageous non-knowing are important but do not replace the entry's mission, authority, measurement, cooperation, and governance center.¹⁴

Score 1 marks knowledge, expertise, and professional autonomy and strategy, competition, and adaptation. Operational, administrative, forensic, and legal knowledge mattered, as did competition among units and adaptation after scrutiny, but the evidence does not support broad professional autonomy or a general strategic-performance account.¹⁴

Score 0 leaves work design, productivity, and automation and innovation, entrepreneurship, and renewal outside the defining comparison. Zero does not mean that work was absent or methods never changed. It means that productivity and innovation language would obscure criminal agency, victimization, and the already better-grounded concepts.¹⁴

The profile is editorial comparison of the cited narrative. Military-security and state-bureaucratic authority span central command, rules, mission command, and frontline action. Hierarchy, metrics, doctrine, training, and teams coordinate; knowledge can move top-down, bottom-up, and through specialist staff; measurement is operational, behavioral, and mission-oriented. After-action review, apprenticeship, doctrinal revision, local iteration, central reconfiguration, and crisis mobilization describe bounded mechanisms, not organizational merit.¹⁴

Members are the sole named beneficiary group because the record supports internal leave, pay, recognition, evaluation, promotion, and status benefits without showing that the state, public, or people owed protection benefited from the criminal system. The member impact remains mixed because other members faced pressure, retaliation, blocked careers, or criminal exposure. Metric gaming, suppressed voice, leader dependence, externalized harm, and capture are comparative risk labels, not measured prevalence or excuses for conduct.¹⁴

Benefits and burdens do not fit inside the combat ledger

Workers and members cannot be aggregated into one culpability or welfare category: some committed or enabled crimes and obtained benefits, while others resisted, faced pressure, cooperated, or carried professional and legal risk. Customers and users of state protection could be converted into targets. Recruiters, informers, transporters, and paramilitary actors appear as alleged, testified, or adjudicated participants in particular records, but “supplier or partner” does not confer voluntariness, legality, or equal responsibility.³⁵¹²

The public military form has no firm-like owner or investor class, so that subject remains research-needed despite the Commission's finding that public funds and goods were misused.¹⁰ Public institutions carried investigative, forensic, judicial, missing-person, truth, and reparative obligations. Mission-beneficiaries were burdened because people owed protection were killed, disappeared, and recoded as threats. Families and communities carried searches, stigma, health and financial costs, threats, displacement, and damaged trust while also building the counter-record.⁷⁹

Nonhuman-life and ecosystem directions remain unclear because the located source set does not assess them systematically; graves and military material do not by themselves establish an ecological effect or attribution. Future generations bear a documented but unquantified burden through children's grief, disrupted education and family plans, economic loss, stigma, distrust, and unresolved identification and accountability work.¹⁵

Paths into deeper study

• Build a case-stage ledger that keeps reports, allegations, charges, acknowledgments, first-instance findings, appeals, final judgments, sanctions, identifications, exhumations, and reparations distinct without exposing victims or witnesses.
• Obtain and preserve family- and victim-organization-controlled records under their access, privacy, correction, attribution, and governance terms rather than treating official mediation as the only public voice.
• Test command and control case by case against operation orders, radio and meeting records, personnel evaluations, rewards, forensic files, unit transfers, warning signs, refusal, investigation, and adjudicated knowledge; do not infer liability from rank or aggregate victim counts.
• Compare security measures that track civilian protection, lawful capture, territorial access, avoided harm, community trust, demobilization, and independent verification without recreating one high-powered target.
• Audit public funds, opportunity costs, cemetery and land effects, animals, ecosystems, and intergenerational health, education, livelihood, and trust with explicit attribution boundaries.