TRACE: An Interactive AI and Compliance-by-Design Framework for Financial Crime Detection

The use of artificial intelligence is growing in the domain of cyber-enabled financial crime detection and has created a significant accountability gap: AI models have been increasingly influencing the regulatory compliance outcomes, and yet their logic is opaque to the humans who are experts and bear legal responsibility for decisions (International Monetary Fund, 2026; Radončić, 2026). Instead of designing AI models to be transparent from their foundation, institutions add the layer of post-hoc explanations onto the autonomous or semi-autonomous AI models (European Commission, 2021; FATF, 2021). Recent research on interactive AI shows analysts prefer to interact and be actively involved with the systems rather than passively consume the outputs (Raees et al., 2024; Wang, 2019). This paper argues that, in high-risk domains such as anti-money laundering, cyber-enabled financial fraud, and cybercrimes, AI should function as a rule-based and human-in-the-loop decision-support system which structures evidence and applies typologies, rules, and auditable variables to support risk assessments while retaining human judgement for final decisions (Bank for International Settlements, 2025; Fink, 2025). This paper introduces TRACE (Transaction Risk Analysis and Compliance Engine) as an interactive AI; it is the manifestation of the compliance-by-design principle. TRACE integrates a transparent rule-based support engine and mandatory human analysts’ approval into a single compliance-grade architecture. In this design, the AI only suggests the risk scoring and flags up factors. To provide support for analysts to make decisions without compromising on transparency, TRACE integrates an auditable retrieval module that draws upon the curated regulated knowledge base, such as FATF, RBI guidelines, and national KYC derivatives. Upon analyst request, contextual evidence, and surfacing retrieved snippets with reliable sources, it remains separate from rule-based scoring or risk narratives (Johnson et al., 2021; Lewis et al., 2020; Saison Technology, 2026). The grounds on which TRACE operates are using the foundational principles of ethics and responsible AI. The risk alerts are flagged using transparency and precise regulatory standards with the aim of mitigating demographic bias in compliance assessments. As it strictly functions as a compliance support tool and not as any enforcement system, it reflects a clear commitment towards responsible AI in this domain, where errors carry serious consequences (AML RightSource, 2024; European Commission, 2021).