Published November 17, 2026 | Version 1

ARTIFICIAL INTELLIGENCE AGAINST CORRUPTION: A Proposal for an Automated and Independent System of Investigation and Justice

Description

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.

Files

Marzal - AI vs Corruption.pdf

Files (664.9 kB)

Name Size
md5:704915ef9ed85e3d249f445d72930d1d
664.9 kB Preview Download

Additional details