Navigating the Disclosure-Security Paradox in Financial Crime Detection: AI, Due Process, and the Crisis of Algorithmic Evidence

The integration of artificial intelligence into financial crime detection has created a fundamental legal paradox that sits at the heart of contemporary crises in justice, security, and technological governance. Constitutional due process requires defendants to meaningfully challenge the evidence used against them, yet disclosing how AI systems identify suspicious activity equips sophisticated criminal actors to evade detection. As algorithmic evidence shifts from a peripheral investigative tool to foundational proof of guilt, courts, regulators, and financial institutions are navigating profound uncertainty about how to reconcile transparency with security. This paper analyses the inadequacy of existing frameworks: the US Brady doctrine, the UK Criminal Procedure and Investigations Act 1996, and the EU AI Act in resolving this tension. Drawing on emerging jurisprudence from State v. Loomis through State v. Arteaga, the UK Fisher Review on Disclosure (2025), and lessons from the Post Office Horizon scandal, it identifies a proportionality principle in which constitutional demands for disclosure scale with the centrality of AI evidence to determinations of guilt. Building on this analysis, the paper proposes a three-tier graduated disclosure framework: minimal disclosure for peripheral AI evidence, intermediate disclosure for substantial AI evidence, and full disclosure where algorithmic outputs are foundational to proof of guilt. The framework is supported by institutional reforms in judicial education, court-appointed expertise, and certification standards. It contributes to sustainable AI governance by ensuring that efficiency in crime detection does not erode the institutional legitimacy that fair trial rights demand.