ARTIFICIAL INTELLIGENCE AGAINST CORRUPTION
A Proposal for an Automated and Independent System of Investigation and Justice
https://zenodo.org/records/24301447
Table of Contents
1. Introduction: The Institutional Problem of Prosecuting Corruption 5
1.1. International Evidence and Contributions of the Series 5
2. From Institutional Discretion to Impunity 6
2.1. Discretion, Opacity and Accountability 6
2.2. Impunity as a Mechanism of Reproduction 7
2.3. Loss of Credibility of Legal Protection 7
3. The Proposal: An Automated Artificial Intelligence System Against Corruption 8
3.2. Institutional Architecture and National Delegations 8
B. Civil Reprisal Proceedings 9
4. Operation of the Fully Automated Procedure 9
4.2. Admission and Investigation 10
4.4. Anti-Corruption Decisions 11
4.5. Human Judicial Appeals 11
5. Enforcement and Monitoring of Decisions 11
6. Encouraging Criminal Complaints Concerning Abuses in Other Jurisdictions 12
7. Governance and Institutional Independence 12
7.2. Prohibition of Control by Public Officials 13
7.3. Prohibition of Control by Legal Professionals 13
7.4. Participation by Intergovernmental Organizations 13
7.5. Technical Independence 14
8. Financing and Sustainability 14
9. Presumptions and New Principles for Investigating Corruption 15
9.1. Presumption of Institutional Corruption Risk 15
9.2. Provisional Legal Presumption of Individual Guilt During Investigation 15
9.4. The Presumption of Guilt in the Public and Institutional Sphere 17
9.5. Legal Duty to Conduct an Exhaustive Investigation 17
9.6. Prospective Investigations and Discovery of Related Offences 18
9.7. Enhanced Confidentiality and Defense Safeguards 18
9.8. Comparative Legal Precedents and Limits of the Proposal 19
10. Legal Feasibility and Relationship with the International Framework 20
11. Spanish Documentary Background: Illustrative Value and Evidentiary Limits 21
12. Progressive Implementation and Evaluation 22
14. Discussion: From Reforming Institutions to an Institutional Alternative 23
Author’s Experience and Accumulated Knowledge 25
Declaration on the Use of Artificial Intelligence 26
Bibliographic and Documentary References 27
Abstract
Three decades after the adoption of the principal international anti-corruption instruments, assessments by intergovernmental organizations continue to document regulatory deficiencies, non-compliance, shortcomings in enforcement and limited results. The first part of the author’s research series systematizes this evidence, analyzes its macroeconomic and systemic consequences, examines existing definitions of corruption, and proposes a new definition and an integrated analytical framework.
This paper raises a fundamental institutional question: can a system for prosecuting corruption be effective when its operation depends on public officials who retain broad powers to decide which complaints to investigate, which investigative measures to undertake and which actions to pursue or abandon, particularly when complaints concern the institutions themselves?
The central hypothesis is that discretion, opacity and inadequate institutional accountability can facilitate impunity, which in turn may become a mechanism for reproducing systemic corruption.
As an alternative, the paper proposes a specialized, fully automated artificial intelligence system capable of receiving complaints, investigating, conducting proceedings, examining the parties’ submissions and evidence, issuing judicial decisions and supervising their enforcement. Its initial jurisdiction would encompass criminal proceedings concerning corruption, related offences and criminal reprisals, including those arising from possible judicial misconduct in other jurisdictions, as well as civil reprisal proceedings involving whistleblowers and persons reporting corruption, including journalists and public officials.
The model envisages plural governance in which public officials and their representatives can never hold a majority or exercise effective control, independent financing partly linked to recovered assets, and mechanisms for human judicial review.
The paper sets out two complementary innovations: an institutional one, through a specialized, automated and independent jurisdiction; and a legal-procedural one, through a provisional legal presumption of individual guilt limited to corruption investigations, a legal duty to investigate exhaustively and an enhanced confidentiality regime. The former does not necessarily depend on adoption of the latter. A separate presumption of institutional risk is also distinguished, which does not attribute guilt to the institution’s members.
The objective is not to introduce artificial intelligence as an auxiliary tool for existing institutions, but to examine the feasibility of an institutional alternative intended to reduce impunity, restore the credibility of legal protection and strengthen the rule of law.
1. Introduction: The Institutional Problem of Prosecuting Corruption
The adoption of the 1996 Inter-American Convention Against Corruption, the 1999 Council of Europe instruments and the 2003 United Nations Convention against Corruption represented a fundamental advance in international recognition of the problem.
However, the existence of laws, specialized agencies, national strategies and evaluation mechanisms has not guaranteed effective prosecution of corruption.
Assessments conducted by international organizations over recent decades have identified persistent deficiencies in prevention, criminalization, implementation and compliance with anti-corruption commitments. They have also documented significant problems concerning perceptions of corruption and institutional trust.
The first part of the author’s research series examines these issues in seven articles devoted to macroeconomic consequences, systemic corruption, criminalization gaps, shortcomings in rules and enforcement, anti-corruption outcomes, limitations of existing definitions, and actors who may facilitate corruption. It also advances a new definition and analytical framework.
The present paper builds on this documentary foundation but addresses a different problem: designing an institutional alternative.
The fundamental question is:
If the prosecution of corruption depends exclusively on officials who can decide what to investigate and what not to investigate, does society not need an alternative mechanism capable of operating without that dependence?
The question is particularly important when complaints concern senior public officials, members of the institutions responsible for investigation or adjudication, or networks of interests capable of influencing them.
In such circumstances, concentrating investigative and decision-making powers in institutions potentially affected by conflicts of interest poses a structural problem.
This paper does not presume that all officials act corruptly or that every decision to dismiss a case or acquit is illegitimate. Rather, it argues that institutional architecture must be capable of functioning even when particular officials, bodies or networks of power have been compromised.
1.1. International Evidence and Contributions of the Series
The author’s series distinguishes three levels that must not be conflated: the magnitude and consequences of corruption, gaps in the regulatory system, and the results of its implementation. Article 1 examines international estimates of macroeconomic costs and public debt; Article 2 studies institutional transformation and systemic corruption; Article 3 identifies the criminalization gap; Article 4 distinguishes normative, compliance and enforcement gaps; Article 5 assesses outcomes; Article 6 reviews seven limitations of corruption definitions; and Article 7 examines the mechanisms for selecting and reproducing corrupt actors and extractive minorities [M1–M7].
European Union findings illustrate the credibility problem: Special Eurobarometer 573 (2026) reports that 71% of respondents consider corruption widespread in their country, 69% believe high-level corruption cases are not sufficiently prosecuted, and only 35% consider government efforts against corruption effective. Furthermore, 54% say they would not know where to report it [O1]. These are respondents’ perceptions, not direct measurements of offences or proof of judicial corruption.
Regarding protection of the EU’s financial interests, the European Public Prosecutor’s Office reported 2,666 active investigations at the end of 2024, with estimated damage exceeding EUR 24.8 billion; approximately 3% of the offences investigated concerned active or passive corruption of public officials [O2]. Without knowing the universe of offences and jurisdictional mandates, this proportion does not establish that officials are insufficiently investigated. OLAF, meanwhile, closed 246 investigations in 2024 and recommended recovery of EUR 871.5 million; a recovery recommendation does not amount to recovery actually carried out [O3].
These figures demonstrate institutional activity while also showing the need to distinguish complaints, investigations, charges, convictions, recoveries and effective whistleblower protection. The series argues that counting laws, agencies and activities is insufficient to evaluate reductions in corruption [M4–M7].
Article 6 proposes a structured definition encompassing acts and omissions, abuse of power and functions, violations of law, and undue benefits or harms affecting persons, institutions or systems [M6]. This definition serves an analytical purpose; it does not replace the statutory definition of offences or their evidentiary requirements.
2. From Institutional Discretion to Impunity
2.1. Discretion, Opacity and Accountability
Administrative and judicial procedures confer discretionary powers on the authorities responsible for handling them.
These powers may be necessary to interpret rules, assess evidence and organize resources. Problems arise when they permit investigations to be omitted, relevant requests to be disregarded or inadequately reasoned decisions to be made without effective oversight.
In corruption matters, discretion can have particularly serious consequences because the matters under investigation may involve those exercising public power or people connected to them.
Moreover, when used to favor private interests contrary to the public interest, institutional discretion can contribute to systemic harm. The relationship between corruption, social decay, economic decline and conflict nevertheless requires specific causal analysis and cannot be presumed in every case.
Opacity makes the reasons for certain decisions difficult to ascertain. Inadequate accountability means that errors, omissions or unlawful actions do not always lead to proportionate consequences.
Together, these factors can facilitate impunity.
2.2. Impunity as a Mechanism of Reproduction
Impunity should not be regarded solely as an outcome of corruption.
It can also function as one of the mechanisms through which corruption reproduces itself.
When perpetrators of corrupt conduct perceive that complaints can be ignored, investigations blocked and decisions left unenforced, their perceived risk of legal consequences declines.
A possible causal sequence therefore emerges:
Discretion, opacity and lack of accountability → omitted or inadequate investigations → impunity → reduced perceived risk of sanction → reproduction and possible expansion of corruption.
This sequence is an explanatory hypothesis requiring empirical testing. It does not imply that all institutional settings follow the same mechanism or that international assessments have independently established every causal link.
Nevertheless, it identifies a limitation of conventional reforms: expanding an institution’s powers or resources does not necessarily resolve the problem when the institution itself may be affected by capture, conflicts of interest or inadequate oversight.
2.3. Loss of Credibility of Legal Protection
The inadequacy of anti-corruption mechanisms has a second consequence: weakening the credibility of legal protection.
Administrative remedies and judicial proceedings may cease to offer a reasonable expectation of protection for those who believe they have suffered institutional abuses.
When citizens perceive complaints as costly, lengthy, risky or futile, they may give up defending their rights.
A decision not to report may reflect fear of reprisals, financial costs, procedural complexity or the expectation that authorities will not investigate effectively.
This withdrawal may reduce the detection of offences and reinforce impunity.
Accordingly, the alternative system should be measured not only by numbers of investigations or convictions, but also by its capacity to provide a credible, accessible and verifiable means of legal protection.
3. The Proposal: An Automated Artificial Intelligence System Against Corruption
3.1. General Principle
The proposal is to establish a specialized artificial intelligence system that fully assumes ordinary investigative, procedural and initial adjudicative functions in certain corruption-related proceedings.
Automation is an essential feature of the model.
This is not a proposal to use artificial intelligence to assist prosecutors, judges or investigators while they retain the principal decisions.
Instead, the automated system would take those decisions under the established legal framework, with procedural safeguards, technical controls and human judicial review.
The proposal would require legislative and, depending on the legal system, constitutional reforms. It is not presented as a power currently granted to AI systems under international or European law.
As an initial approach, independent pilot tests could analyze case files and decisions to identify indications of judicial irregularities, detect inconsistencies and produce verifiable assessments. Their findings could help investigate possible offences and improve public information, without automatically equating a legal disagreement with corruption.
3.2. Institutional Architecture and National Delegations
The system would have a common institutional architecture and could establish national delegations adapted to the respective legal systems.
The United Nations Convention against Corruption (UNCAC) would provide a fundamental reference for defining its substantive scope, supplemented by regional instruments and national legislation.
National delegations would incorporate relevant offences under each legal system through legitimate legislative processes and in compliance with the principle of legality in criminal law.
For example, several countries would need expressly to consider misconduct in public office involving knowingly unlawful decisions and other offences committed by public officials in the performance of their duties.
The system could neither create offences retroactively nor impose penalties for conduct not criminalized under the applicable law.
3.3. Initial Jurisdiction
At the outset, jurisdiction would be limited to two broad areas.
A. Criminal Proceedings
These would cover corruption offences and related crimes within the system’s jurisdiction, especially those allegedly committed by public officials or connected with abuse of institutional power.
They would include possible offences arising from actions taken in proceedings before other jurisdictions, provided sufficient indications existed.
An adverse judicial decision, even if subsequently overturned, would not by itself constitute proof of judicial misconduct or any other offence.
They would also include criminal proceedings concerning reprisals against whistleblowers and persons reporting corruption, including public officials and journalists who may be important sources for exposing corruption.
B. Civil Reprisal Proceedings
These would cover civil proceedings involving reprisals against whistleblowers, reporting persons, journalists and other persons protected because of their involvement in reporting, investigating or lawfully disclosing possible corruption.
Jurisdiction would require a legal determination of the connection between the proceedings and the alleged reprisal.
The purpose would be to prevent certain procedural instruments from being used to obstruct corruption reporting or intimidate those investigating it.
4. Operation of the Fully Automated Procedure
4.1. Filing Complaints
The system would provide accessible channels for filing complaints and submitting documents.
It should accept documentary evidence, judicial decisions, administrative records, communications, testimony and other legally relevant material.
Submission and registration would not depend on prior acceptance by a public official; preservation, integrity and traceability measures would be introduced to prevent complaints from being lost, altered or destroyed.
Confidentiality and data protection mechanisms would be guaranteed, along with assisted access for people facing technological barriers.
4.2. Admission and Investigation
The AI would examine a complaint and determine whether the legal requirements for opening an investigation had been met.
Decisions to reject or close cases would have to be reasoned, communicated to the participants and subject to effective challenge, both within the system and before existing courts.
Where an investigation was warranted, the system would identify appropriate investigative steps and issue legally authorized requests.
Material actions that cannot be automated would be performed by the competent services, without thereby granting those services discretionary powers to determine the outcome.
Coercive measures or restrictions on fundamental rights could require indispensable human authorization and oversight for as long as the applicable legal framework so provides.
In the legally defined circumstances described in Section 9, opening an investigation would trigger a duty of exhaustive verification, including assessment of other potentially related corruption offences and a search for exculpatory evidence. Decisions to broaden an investigation, refuse investigative steps or close a case would be reasoned and reviewable. Where appropriate, enhanced confidentiality would apply temporarily and proportionately.
4.3. Hearing the Parties
A fundamental principle would be to ensure that all affected persons can be effectively heard, even without formal participation as a party where the law permits, distinguishing the right to provide information from procedural powers reserved to persons with legal standing, without impairing the rights of the defense.
The AI would have to record, examine and respond to relevant submissions, evidence and procedural requests.
It would not be sufficient to produce a formally reasoned decision that omitted essential issues raised during the proceedings.
The system should make it possible to verify which requests were submitted, how they were handled and why they were accepted or rejected.
Proceedings declared confidential from the public or, exceptionally, from a party would not eliminate the right to be effectively heard or access to the essential material needed to challenge restrictive measures. Temporary restrictions would be documented and subject to independent review.
Automation would not justify reducing defense rights, adversarial safeguards, access to the case file or legal assistance; any procedural adaptation would require justification and equivalent safeguards.
4.4. Anti-Corruption Decisions
The AI would issue the relevant decisions, including orders and judgments of acquittal or conviction, within the powers granted by legislation that might eventually implement the model.
Decisions would identify the facts found proven, relevant evidence, applicable rules and legal reasoning employed.
Their reasoning would have to be verifiable and reviewable, rather than merely an apparent explanation generated by the system.
In criminal matters, granting the power to impose custodial sentences would require particularly rigorous constitutional and human-rights justification.
Where it identified conflicting case law, the system would flag it and activate legally established mechanisms for review or harmonization, without assuming powers outside its jurisdiction.
4.5. Human Judicial Appeals
As with current decisions of courts of first instance, the system would provide for appeals to higher judicial bodies.
Review would permit examination of facts and law under the applicable safeguards, correction of errors and annulment of decisions affected by technical or procedural defects.
Where a higher court modified or overturned an automated decision, the AI could prepare a reasoned assessment accessible to the parties explaining the grounds for the decision and available avenues of appeal. Only if specific and sufficient indications of judicial misconduct or another offence emerged would opening criminal proceedings be considered; mere disagreement or reversal would not suffice.
The availability of an appeal would not automatically resolve all constitutional objections to an initial AI decision.
The proposal must therefore analyze which safeguards require prior human involvement and which could be satisfied by subsequent controls.
5. Enforcement and Monitoring of Decisions
The effectiveness of legal protection does not end when a decision is issued.
A legally well-founded decision may be ineffective if it is not enforced or if enforcement is unjustifiably delayed.
The proposed system would assume automated monitoring of enforcement of its decisions.
Court offices and other legally competent services would perform the necessary material actions but would be subject to verifiable requirements and a compliance monitoring system.
The AI would record deadlines, request information about steps taken, detect non-compliance and trigger the procedural consequences provided by law.
The parties could monitor enforcement status and report relevant omissions.
The aim is to prevent non-enforcement from becoming another route to impunity.
This oversight should also apply to the automated system itself: its failures, delays or breaches would have to be identifiable and correctable.
6. Encouraging Criminal Complaints Concerning Abuses in Other Jurisdictions
The system should not merely wait passively to receive complaints.
Its implementation would include an information strategy to help victims of possible institutional offences understand and exercise their legal rights.
Particular attention would be paid to actions taken in administrative, civil, labor and administrative-law proceedings where there were indications of corruption offences, judicial misconduct or other conduct falling within the system’s criminal jurisdiction.
Encouraging complaints would not mean presuming that every harmful institutional action is criminal.
The objective would be to enable victims to report potentially criminal acts without depending on the same institutional mechanisms whose conduct they challenge.
Anti-corruption and whistleblower-protection organizations would play an important role in informing, advising and protecting affected persons, as well as improving the system.
The system’s credibility would be decisive.
If victims could reasonably trust that their complaints would be examined, their evidence assessed and decisions enforced, some of the barriers currently discouraging reporting would diminish.
An increase in complaints should not automatically be interpreted as an increase in corruption: it could also reflect improved detection.
7. Governance and Institutional Independence
7.1. Plural Governance
The system would require a governance structure with clearly defined functions.
Participation is proposed for organizations specializing in anti-corruption or whistleblower protection, other civil-society representatives, experts in law, technology and auditing, and representatives of judges and prosecutors.
Transparency International’s Advocacy and Legal Advice Centres (ALACs) could provide a useful organizational reference because of their experience with corruption and advice and support for reporting persons.
Their possible participation would require explicit agreements and cannot be presumed.
7.2. Prohibition of Control by Public Officials
A fundamental institutional rule is established:
Public officials and their representatives must never constitute a majority or exercise effective control, directly or indirectly, over any governing, supervisory or auditing body of the system.
This restriction also covers indirect means of control through appointments, affiliated entities, voting agreements or veto powers.
Representatives of judges and prosecutors may participate and contribute professional expertise, but must not control the system.
At the same time, majority participation by non-governmental organizations must not become a new form of institutional capture.
Controls will be necessary regarding financing, conflicts of interest, representatives’ independence and relationships with investigated entities.
No governing body may modify decisions in individual cases outside the legally established review mechanisms.
7.3. Prohibition of Control by Legal Professionals
The prohibition on majority status and effective control would extend to legal professionals as a whole—including legal academics, lawyers, prosecutors and judges—and their organizations, counted collectively. This rule seeks to prevent the replacement of one institutional dependency by control exercised by a single professional group.
7.4. Participation by Intergovernmental Organizations
Intergovernmental organizations could play a significant role in this system.
Their experience in standard-setting, evaluation and international cooperation could help advance the system and even justify limited participation in its governing bodies.
Participation by intergovernmental organizations would be limited and counted together with the representation of officials and legal professionals to prevent control by these groups. Independent civil society, domestic and international, would retain an effective voting majority, without prejudice to mechanisms preventing private capture.
7.5. Technical Independence
Independence must also extend to the technological infrastructure.
The system cannot depend on a supplier, administrator or group of specialists with uncontrolled capacity to modify its operation.
Version records, external audits, integrity controls, safeguards against manipulation and mechanisms for suspending operations when serious failures are detected would be required.
Fully automating the procedure does not eliminate the responsibility of the people and organizations developing, maintaining and supervising the infrastructure.

8. Financing and Sustainability
Initial public financing is proposed, contributed by participating states for at least three years, with guarantees of adequate funding and no ability to interfere in proceedings.
If appeals to higher courts were resolved without undue delay, recovered assets could contribute significantly to financing the system. The possibility of achieving self-sufficiency quickly is an economic hypothesis requiring testing against recovery scenarios, costs and the legal availability of funds.
During the first ten years, the proportion allocated to financing could be higher to facilitate technological development, the establishment of national delegations and expansion of the system.
After that period, a lower percentage would be established for system maintenance and upgrades.
The mechanism would require an independent fund subject to public and external audits.
Financing could come only from legally available resources after respecting victims’ rights, restitution obligations and international asset-recovery agreements.
The model should avoid any financial incentive to issue convictions or confiscation orders.
Neither individual decisions nor the income of those administering the system could depend on the monetary value of penalties imposed.
9. Presumptions and New Principles for Investigating Corruption
9.1. Presumption of Institutional Corruption Risk
The first proposal is to examine a presumption of institutional risk based on objective indicators of systemic corruption, repeated non-compliance and documented deficiencies in oversight.
Its function would be to trigger enhanced measures for transparency, investigation, whistleblower protection and supervision.
It would not amount to declaring the members of an institution guilty.
The presumption should be based on public, verifiable criteria open to challenge.
9.2. Provisional Legal Presumption of Individual Guilt During Investigation
As a deliberately controversial legal reform, the proposal is that, where individualized objective indications of possible corruption offences exist, a provisional legal presumption of guilt should operate during the investigative phase. Unlike a mere police hypothesis, the proposal seeks expressly to displace, only during this phase and for its specific purposes, the legal presumption of innocence as the rule guiding the investigation.
The trigger would not be mere membership of an institution, holding an office or a general finding of systemic corruption. There would have to be verifiable facts or indications concerning a person and conduct, subject to review and challenge. However, the investigation would not have to have already obtained the very evidence it is intended to seek.
Its specific legal effect would be to activate an enhanced duty to investigate and verify indications, including exculpatory evidence. It would not automatically authorize arrests, searches, interceptions, asset freezes, penalties, reversal of the burden of proof at trial or convictions. Measures restricting rights would retain their independent legal prerequisites and controls.
The presumption would be temporary, individual, reasoned and reviewable. It would cease when objective indications disappeared, a duly justified decision to close the case was made or the investigative phase ended. Trial and conviction would remain subject to the presumption of innocence and the duty to prove criminal responsibility to the applicable standard.
The proposal expressly acknowledges its conflict with constitutional and treaty-based protection of the presumption of innocence throughout criminal proceedings. Its adoption would require far-reaching legal reforms and an examination of necessity, proportionality and compatibility with international human-rights instruments.
The aim is to prevent the absence of evidence that could only be obtained through investigation from becoming a reason not to investigate, particularly where institutional networks may be involved and structural barriers hinder access to documents, testimony or financial information.
The provisional individual presumption and the presumption of institutional risk are distinct: the former concerns personal indications of criminal conduct; the latter identifies risks in institutional functioning and triggers supervisory measures without individual attribution of guilt.
Implementation of an automated investigation and justice system would not necessarily depend on adoption of this legal exception; the two reforms must be assessed separately.
9.3. Pretrial Detention: An International Precedent and the Tension Between Formal Presumption and Material Effects
Pretrial or provisional detention is a widespread institution in contemporary criminal justice systems. Subject to conditions that differ among jurisdictions, it allows a person to be deprived of liberty before conviction, on the basis of indications or a provisional assessment of criminal responsibility and additional precautionary aims, such as preventing flight or interference with evidence.
From a material and critical perspective, this anticipatory assessment may be interpreted as a provisional attribution of possible guilt with actual legal consequences, even though legal systems formally maintain that the presumption of innocence remains intact. The author questions the coherence between that formal assertion and deprivation of liberty before conviction.
The hypothesis that formal preservation of the presumption of innocence also serves an institutional legitimizing function must be critically examined; it is not an established uniform motive of legislators and courts in all countries. Prevailing doctrine distinguishes an evidentiary assessment of possible responsibility, precautionary purposes and a legal declaration of guilt.
The analogy raises a proportionality question: if measures as intrusive as pre-conviction deprivation of liberty may be permitted, why exclude in advance a provisional presumption intended only to trigger a rigorous corruption investigation? The comparison does not itself establish the legal compatibility of formally displacing the presumption of innocence, and an investigation may entail highly intrusive steps requiring specific safeguards.
9.4. The Presumption of Guilt in the Public and Institutional Sphere
The provisional presumption of guilt is not a phenomenon confined to criminal proceedings. In the public sphere, disclosure of possible acts of corruption may lead to the resignation, dismissal or suspension of those implicated, even before any judicial judgment has been issued. These consequences may arise from voluntary decisions, demands for political accountability, precautionary measures or premature judgments of guilt.
This phenomenon highlights a difference between the legal presumption of innocence and the material consequences that may follow indications of responsibility. While criminal proceedings formally maintain that presumption, certain institutions and social actors make decisions that anticipate adverse consequences for the person concerned.
The proposal calls for examining this divergence and determining whether it is legally coherent to establish an express, regulated provisional presumption during the investigative phase, subject to individualized objective indications, defense safeguards and review mechanisms, without altering the standard of proof required for a criminal conviction.
A resignation does not necessarily establish that the person concerned is regarded as guilty: it may reflect a loss of confidence, public pressure or an attempt to protect an institution. The relevant observation is that adverse institutional consequences can precede a judicial finding of guilt. Social or institutional presumptions of guilt must therefore be distinguished from the provisional legal presumption proposed in this paper.
9.5. Legal Duty to Conduct an Exhaustive Investigation
The second legal-procedural innovation, complementary to rather than an alternative to the provisional presumption, is a legal duty to investigate exhaustively where objective indications of corruption exist. Its rationale is to prevent discretion over admission, selection of investigative measures or case closure from prematurely terminating cases in which evidence requires further investigation.
The duty would encompass verification of reported facts; searches for relevant documents and testimony; examination of explanations and exculpatory evidence; identification of other possible corruption offences committed by the same person under investigation; and investigation of possible participants, beneficiaries, persons concealing offences or related organized structures.
The system would record all relevant evidentiary requests by the parties and the reasons for undertaking or refusing each measure. Closing a case would require verifiable reasons addressing the indications examined, evidence obtained, reasonable alternatives and why further action was unwarranted. Decisions would be appealable.
Exhaustiveness does not mean undertaking every conceivable measure or conducting an indefinite investigation: relevance, proportionality, fundamental rights, reasonable time limits and independent oversight would continue to apply. Exculpatory material must also be documented and actively sought.
9.6. Prospective Investigations and Discovery of Related Offences
Corruption may operate through concealed conduct and networks not fully revealed in the initial complaint. The proposal therefore calls for reviewing, in legal systems that impose them, restrictions on prospective investigations, so that objective indications of corruption allow examination of other possible offences committed by the person already under investigation and links with third parties or institutional structures.
Expansion must be grounded in verifiable indications or patterns and a reasonable connection to the corruption under investigation. It does not authorize indiscriminate searches for violations, general surveillance of the population or unrestricted access to private data. Every intrusive measure would require the corresponding legal basis, reasoning and authorization.
AI could identify patterns, documentary contradictions, temporal coincidences and relevant relationships, but every inference would have to be checked against original data and alternative hypotheses considered. Decisions to broaden the scope of an investigation would be recorded, reasoned and reviewable.
9.7. Enhanced Confidentiality and Defense Safeguards
The system could establish different confidentiality levels for investigations whose publicity might compromise evidence, facilitate interference, endanger whistleblowers and witnesses or cause disproportionate reputational damage. Some expanded or prospective investigations might require enhanced confidentiality functionally comparable to investigative secrecy orders, without automatically transplanting the rules of any one national legal system.
Two autonomous legal reforms must also be distinguished: granting jurisdictional powers to AI systems and recognizing a provisional legal presumption of guilt during corruption investigations. The former does not automatically legitimize the latter. The presumption of innocence is protected throughout criminal proceedings by international instruments, including Article 14(2) of the International Covenant on Civil and Political Rights, Article 6(2) of the European Convention on Human Rights and, in the EU, Article 48 of the Charter and Directive (EU) 2016/343. The proposed exception therefore faces particularly profound compatibility objections.
A distinction must be drawn between confidentiality vis-à-vis the public, protection of sources and sensitive data, and exceptional restrictions on a party’s access to particular proceedings. The latter requires particularly rigorous justification, limited duration, independent review and mechanisms ensuring defense rights and an effective opportunity to challenge harmful measures.
Confidentiality would not be triggered automatically by a provisional presumption of guilt. Every decision would require concrete risks, stated reasons, time limits and traceability. The system would record who accessed which information, for what reason and when confidentiality was lifted.
Public transparency would be ensured through aggregate statistics and audits, publication of general criteria and accountability compatible with confidential case files. External confidentiality must not become internal opacity or a means of concealing irregularities.
9.8. Comparative Legal Precedents and Limits of the Proposal
Among comparative precedents involving evidentiary presumptions is illicit enrichment. Article 20 of the United Nations Convention against Corruption invites consideration of criminalizing a significant increase in a public official’s assets that cannot reasonably be explained in relation to lawful income, subject to the constitution and fundamental principles of each legal system [N1]. This precedent is not equivalent to the proposed investigative presumption.
Lindy Muzila, Michelle Morales, Marianne Mathias and Tammar Berger comparatively examined criminalization of illicit enrichment in On the Take (World Bank/UNODC, 2012), specifically in light of the evidentiary difficulties posed by corruption offences and asset recovery [A1]. This approach does not presume every official guilty; rather, it starts from financial facts that must be established.
A particularly significant statutory precedent is Section 10 of Hong Kong’s Prevention of Bribery Ordinance: it requires an official to provide a satisfactory explanation for resources or a standard of living disproportionate to official income. Its compatibility with the presumption of innocence has been subject to legal debate, including consideration by the United Nations Human Rights Committee [A2]. It must not be extrapolated to judicial misconduct offences or guilt by institutional membership.
In Salabiaku v. France (1988), the European Court of Human Rights recognized that presumptions of fact or law may exist in criminal matters, but required reasonable limits, proportionality and respect for defense rights [J1]. This case law does not authorize indiscriminate reversal of the burden of proving guilt.
The debate must distinguish established underlying facts, rebuttable evidentiary inferences, duties to explain wealth and a provisional legal presumption of guilt limited to investigation. The latter seeks expressly to modify the initial legal rule during a particular phase, without thereby reversing the burden of proving guilt for conviction. The constitutional and treaty-based objection is particularly serious and remains unresolved.
These precedents do not establish the legal admissibility of the proposal. They must be studied alongside international safeguards concerning the presumption of innocence, defense, judicial oversight and effective remedies, without making the automated system dependent on adoption of this exception.
10. Legal Feasibility and Relationship with the International Framework
The proposal affects fundamental principles of judicial organization and procedural law.
UNCAC provides an international reference framework for prosecuting corrupt conduct, but does not itself authorize replacing judges with artificial intelligence systems.
Pilot programs should also measure detection of related offences, the relevance of investigative measures undertaken and refused, identification of exculpatory material, false positives, the duration and review of confidentiality, and the rate at which investigative decisions are overturned. No pilot conducted under existing law could apply an exception to the presumption of innocence prohibited by that law.
However, UNCAC refers to autonomous anti-corruption bodies, and official information provides grounds for presuming that, in most countries, existing anti-corruption bodies cannot be characterized as autonomous.
Similarly, international human-rights instruments recognize guarantees such as the right to a competent, independent and impartial tribunal, the presumption of innocence and the right to an effective remedy.
We face a similar situation: official information calls into question whether independent and impartial courts exist in cases involving corruption and reprisals.
At European level, the Artificial Intelligence Act and rules governing judicial organization and procedure are relevant references for examining applicable restrictions and safeguards.
The experience of the European Public Prosecutor’s Office also allows analysis of how investigative, prosecutorial and judicial oversight functions are distributed.
However, none of these references by itself demonstrates that a fully automated criminal jurisdiction is currently permissible.
The paper seeks precisely to open debate on the reforms that would be required to establish one.
That debate must distinguish three questions: the technical capabilities of AI systems, the legal legitimacy of granting them judicial powers, and the institutional desirability of replacing certain conventional mechanisms.
The investigative presumption could produce confirmation bias, disproportionate inquiries, stigmatization and unjustified expansion of investigations. Enhanced confidentiality could conceal errors or unduly restrict the defense. The design would require verifiable triggering thresholds, active searches for exculpatory evidence, expiry of measures, independent audits, judicial controls and compensation for harm.
11. Spanish Documentary Background: Illustrative Value and Evidentiary Limits
The author has provided three sets of documents: a dossier entitled “COMPLAINTS CONCERNING TAX AUTHORITIES, PROSECUTORS, JUDGES AND POLITICIANS 2002”; a complaint to the General Council of the Judiciary in October 2026 concerning alleged judicial reprisals; and another complaint in November 2026 concerning actions by the Tax Agency and Social Security entities [E1–E3].
The first set lists complaints and documents attributed to tax and judicial proceedings from the early 2000s, with references to accounts, transactions and persons identified by the complainants. No independent audit has been completed of authenticity, currency, duplication, ownership, source of funds, administrative outcomes or criminal proceedings. Accordingly, the overall amount exceeding EUR 5 billion reported by the complainants is not included as an established amount of money laundering.
The October 2026 complaint describes, among other matters, proceedings initiated against persons who had disclosed or reported alleged corruption and challenges procedural decisions concerning admission and closure. The November 2026 complaint requests examination of whether judicial bodies detected and reported indications of offences in proceedings involving tax and Social Security authorities. Both identify verifiable questions: which facts were reported, what evidence was submitted, which investigative steps were taken, what decisions were issued and what review mechanisms existed.
The relevance of the Spanish background to this paper does not lie in attributing criminal liability to identifiable persons without a final judgment, but in illustrating the problem of traceability between tax or administrative complaints, criminal investigation, judicial decisions and protection of reporting persons. An automated pilot system could compare complete case files, locate documentary omissions and procedural inconsistencies, and issue reproducible assessments, without those assessments themselves becoming declarations of guilt.
Academic use of these materials requires, before any final publication containing personal allegations, verification of originals and acknowledgments of receipt, administrative outcomes, judicial decisions identified by ECLI and, where appropriate, the status of proceedings and accounts of the persons concerned.
12. Progressive Implementation and Evaluation
The proposed transformation requires experimental and gradual implementation.
A first stage could use historical case files and simulated proceedings to assess the system’s ability to identify relevant facts, apply legal rules, examine evidence and produce reasoned decisions.
A second stage could permit use in real proceedings under an experimental regime that respects existing safeguards, without initially granting judicial powers incompatible with applicable law. Even its decisions could be of considerable assistance to parties, prosecutors and judges.
Full conferral of powers would require the corresponding legal reforms and independent evaluation of the preceding stages.
Evaluation should consider at least the quality of decisions, protection of procedural rights, errors, reversals, duration of proceedings, effective enforcement, asset recovery and whistleblower protection.
Accessibility, public trust and willingness to report should also be measured.
Credibility must not be presumed merely because the system is automated. It must be demonstrated through verifiable outcomes.
13. Risks and Limitations
The proposal entails significant risks that must be examined openly.
As distinct legal-procedural innovations, the proposal combines a presumption of institutional corruption risk; a controversial provisional legal presumption of individual guilt that would displace the presumption of innocence only during investigation; a legal duty to investigate exhaustively, including related offences and structures; and proportionate levels of enhanced confidentiality. Pretrial detention offers an international analogy for anticipatory assessments of responsibility, without providing direct legal authorization for the proposed exception.
These procedural innovations are separable from the automated institutional architecture. Their feasibility would require comparative analysis, regulatory reforms, specific safeguards and empirical evaluation, particularly given the tension with the presumption of innocence protected by international law.
Centralization of functions may also create new risks of concentrated power.
Independence from public officials does not automatically guarantee independence from technology companies, private organizations or those responsible for infrastructure.
Likewise, granting criminal-law powers to AI raises fundamental questions about democratic legitimacy, legal responsibility and safeguards for accused persons.
Automation may reduce certain opportunities for human interference, but does not itself eliminate the possibility of corruption, arbitrariness or error.
The design must therefore provide for independent controls, transparency, judicial review, liability for harm, data protection and procedures for suspending operations and correcting failures.
These limitations must neither be omitted nor treated as secondary issues. They form part of the proposal’s feasibility assessment.
14. Discussion: From Reforming Institutions to an Institutional Alternative
Most conventional anti-corruption strategies seek to improve existing institutions through new laws, resources, controls and supervisory mechanisms.
These reforms can yield significant advances.
However, where the problem includes possible capture of bodies responsible for investigation or sanctions, internal reform may be insufficient.
The proposal introduces a structural alternative: replacing ordinary decision-making intervention by those bodies with a specialized automated system subject to verifiable safeguards.
Its potential contribution is not limited to speeding up proceedings.
It aims to reduce opportunities for institutional obstruction, facilitate reporting of possible offences committed within institutions themselves and strengthen effective enforcement of decisions.
It also seeks to change victims’ expectations.
A credible legal avenue may increase willingness to report and thereby improve detection of possible offences.
If the system demonstrated greater investigative capacity, independence and enforcement, it could help reduce impunity and alter incentives favoring the reproduction of corruption.
Nevertheless, these effects are hypotheses requiring testing. The superiority of the proposed system cannot be assumed before evaluation.
15. Conclusions
International assessments over recent decades document persistent shortcomings in combating corruption. The first part of the author’s research series analyzes this evidence, its consequences and the limitations of conventional mechanisms, examines existing definitions, and proposes a new definition of corruption and analytical framework.
This paper argues that discretion, opacity and inadequate institutional accountability can contribute to impunity, which may act as a mechanism for reproducing systemic corruption.
In response, it proposes a fully automated artificial intelligence system initially specializing in criminal proceedings concerning corruption and related offences and in civil reprisal proceedings involving whistleblowers, reporting persons and journalists.
The proposal encompasses investigation, procedural management, judicial decision-making and monitoring of enforcement, with human judicial review and defense safeguards.
Its institutional architecture envisages national delegations, plural governance without a majority or effective control by public officials or legal professionals, participation by anti-corruption organizations and financing partly linked to recovered assets.
It also opens debate on the presumption of institutional corruption risk and a controversial presumption of individual guilt in contexts of systemic corruption.
The fundamental objective is to examine whether an automated alternative could offer more credible, independent and effective legal protection than conventional mechanisms in areas where the latter have demonstrated persistent shortcomings.
The proposal does not presume that artificial intelligence is infallible or that its implementation would be legally possible without far-reaching reforms.
Its contribution is to shift debate from merely improving existing anti-corruption institutions to a more fundamental question:
Can society build a mechanism for prosecuting corruption that does not depend on decisions by the same officials and institutional structures which, in certain cases, may themselves be implicated in the matters requiring investigation?
Answering this requires empirical research, technological experimentation and international legal debate. Yet the magnitude of documented problems justifies examining institutional alternatives that have so far received limited attention.
Author’s Experience and Accumulated Knowledge
This research programme is the result of a cumulative process spanning more than fifteen years of practical experience, observation and research into corruption and the functioning of institutional and anti-corruption mechanisms.
During this period, the author’s direct experience in reporting and communicating information concerning possible cases of corruption to Spanish, European and international institutions was broadened through the establishment and leadership of various civil society initiatives: as president of an association representing people affected by institutional actions, as the sole founder and president of a federation of associations, and as the sole founder and president of an alliance comprising associations and a social movement.
These activities enabled him to become directly involved in situations affecting other individuals and organisations, undertake actions before different institutions, and share projects and experiences with other civil society leaders.
This trajectory produced a progressive and cumulative process of knowledge acquisition.
Practical experience raised new questions that led to further research; this expanded the available knowledge and, in turn, prompted the investigation of other problems, the reconsideration of previously examined issues and the identification of relationships between phenomena that are commonly studied separately.
This process has influenced not only the questions addressed by the research, but also how those questions have been formulated, the sources and data examined, the relationships established between them, and the analytical framework through which the evidence is interpreted.
The resulting research has progressively connected the macroeconomic dimension of corruption with its systemic nature, the gaps between the phenomenon and its criminalisation, the gaps in the functioning of anti-corruption systems, their measurable outcomes, and the mechanisms through which corruption may adapt, persist, reproduce itself and contribute to institutional, economic, political and societal transformations.
The author’s accumulated knowledge has thus helped shape a research programme that goes beyond the study of individual acts of corruption to examine jointly the scale and nature of the phenomenon, the structures within which it operates, and the systems established to combat it.
The specific cases known to the author through this experience do not form part of the empirical evidence used in the seven articles comprising Part I of the series.
Their empirical basis consists of the documentary sources, official data and published evidence cited in each article.
Instead, the author’s experience and accumulated knowledge have contributed to identifying the questions to be asked, selecting the dimensions to be investigated, connecting evidence produced across different institutional and disciplinary contexts, and integrating these elements into the analytical framework developed throughout the series.
Declaration on the Use of Artificial Intelligence
ChatGPT (OpenAI) was used to assist with the drafting, linguistic revision, translation, document structuring and presentation of this research paper.
The research questions, conceptual framework, hypotheses, institutional proposal, analysis, arguments and conclusions are the responsibility of the author, who retains full responsibility for the content, accuracy and integrity of the publication.
Bibliographic and Documentary References
Author’s Research
[M1] Marzal, J. (2026). Without Corruption There Would Be No Public Debt. Macroeconomic Corruption. Corruption Series, Part I, Article 1. https://doi.org/10.5281/zenodo.21994962
[M2] Marzal, J. (2026). Systemic Corruption: From Institutional Consequences to Systemic Transformation. Corruption Series, Part I, Article 2. https://doi.org/10.5281/zenodo.21995042
[M3] Marzal, J. (2026). The Criminalization Gap in Corruption: Impunity May Begin Before Prosecution. Corruption Series, Part I, Article 3. https://doi.org/10.5281/zenodo.22733321
[M4] Marzal, J. (2026). Three Gaps in the International Fight Against Corruption: 30 Years of Rules, Non-Compliance and Inadequate Enforcement. Corruption Series, Part I, Article 4. https://doi.org/10.5281/zenodo.23019185
[M5] Marzal, J. (2026). Thirty Years of Fighting Corruption: What Have the Results Been? Corruption Series, Part I, Article 5. https://doi.org/10.5281/zenodo.23333997
[M6] Marzal, J. (2026). Towards a Structured Definition of Corruption. Corruption Series, Part I, Article 6. https://doi.org/10.5281/zenodo.23684696
[M7] Marzal, J. (2026). Who Corrupts the Corrupt? From the Individual Corruptor to the Selection and Reproduction of Corrupt Actors. Corruption Series, Part I, Article 7. Manuscript reviewed; publication identifier pending verification.https://doi.org/10.5281/zenodo.24102380
[M8] Marzal, J. (2026). Thirty Years of Anti-Corruption Frameworks: A Structural Misalignment Between Definition, Scale and Results. https://doi.org/10.5281/zenodo.19357789
Intergovernmental Organizations and Legal Sources
[O1] European Commission (2026). Special Eurobarometer 573: Citizens’ Attitudes Towards Corruption in the EU. https://data.europa.eu/data/datasets/s3693_105_1_sp573_eng
[O2] European Public Prosecutor’s Office (2025). Annual Report 2024. https://www.eppo.europa.eu/assets/annual-report-2024/index.html
[O3] European Anti-Fraud Office (2025). OLAF Report 2024. https://ec.europa.eu/olaf-report/2024/index_en.html
[N1] United Nations (2003). United Nations Convention against Corruption, Article 20. https://mandates.un.org/mandate/A/RES/58/4
[N2] Organization of American States (1996). Inter-American Convention Against Corruption.
[N3] Council of Europe (1999). Criminal Law Convention on Corruption and Civil Law Convention on Corruption.
[N4] European Union. Regulation (EU) 2024/1689 on artificial intelligence; Regulation (EU) 2017/1939 on the European Public Prosecutor’s Office; Directive (EU) 2016/343 on the presumption of innocence.
[A1] Muzila, L., Morales, M., Mathias, M. and Berger, T. (2012). On the Take: Criminalizing Illicit Enrichment to Fight Corruption. World Bank and UNODC. https://ideas.repec.org/b/wbk/wbpubs/13090.html
[A2] United Nations Human Rights Committee. Examination of Section 10 of Hong Kong’s Prevention of Bribery Ordinance and the evidentiary burden relating to unexplained enrichment. https://docstore.ohchr.org/SelfServices/FilesHandler.ashx?enc=phdEtwDD2MplM4FPfVdehXS48BjJX1iIWh8gIVPJjq0FUan%2FppuH3RxFlCkUX1UlskRmfSEwsz4F41ycX5MpYY3FZBHTdYujXsSRnvyoGzc%3D
[J1] European Court of Human Rights (1988). Salabiaku v. France, Application No. 10519/83. https://hudoc.echr.coe.int/?i=001-57570
Spanish Documentary Background (Allegations, Not Judicially Established Facts)
[E1] Dossier: “COMPLAINTS CONCERNING TAX AUTHORITIES, PROSECUTORS, JUDGES AND POLITICIANS 2002/2004.” Documentation provided by the author.
[E2] Marzal, J. (2026). Complaint to the General Council of the Judiciary concerning reprisals, 29 October. Document provided by the author.
[E3] Marzal, J. (2026). Complaint to the General Council of the Judiciary concerning Social Security and tax authorities, 2 November. Document provided by the author.