Shadow banking, sanctions evasion, and the AI advantage

Key takeaways

1. Global Iranian shadow-banking uncovered: FinCEN identified a complex web of front and shell companies, mainly in the UAE, Hong Kong, and Singapore, moving billions in illicit USD transactions for sanctioned Iranian actors.
2. Traditional compliance falls short: The network’s scale and use of opaque entities show rule-based compliance systems and manual reviews cannot keep up with evolving schemes.
3. AI enhances detection: AI-powered solutions—like entity resolution and behavioral analytics—can uncover hidden relationships and suspicious patterns that legacy systems miss.
4. Improved efficiency and risk prioritization: AI automates investigations, adapts detection thresholds, and enables proactive risk management, helping institutions reduce false positives and anticipate future evasion tactics.
5. Sensa Risk Intelligence transforms compliance: SymphonyAI’s Sensa Risk Intelligence (SRI) platform uses modular design with agentic AI to automate and optimize compliance, turning it into a strategic advantage rather than just a defensive measure.

How SensaAI is reinventing compliance in an era of complex risk

FinCEN’s recent Financial Trend Analysis (FTA)[1] details a multi-billion-dollar “shadow-banking” network that enables Iran to evade sanctions. FinCEN analyzed suspicious activity reports (SARs) covering transactions from January-December 2024 that showed this shadow-banking network consists of a global web of front companies, shell entities, and exchange-houses that sanctioned Iranian actors use to access the U.S. dollar (USD) system. Additionally, they move funds internationally, and channel resources into illicit activities including oil exports, weapons procurement and financing regional proxies.

Oil and shell companies facilitated most of the money movement

FinCEN identified roughly $9 billion in transactions appearing to relate to Iranian shadow-banking networks in the dataset. Several different types of companies fuel the network, but oil and shell companies were involved in the majority of activity. Approximately 44% of the funds (about $4 billion) were connected to oil company activity. These were primarily foreign entities (from UAE, Singapore) linked to sanctioned Iranian oil. Meanwhile, shell companies (entities with little or no verifiable business activity, often created solely to move money) moved roughly $5 billion, or 56% of total funds. These shell companies frequently originated from Hong Kong and non-resident accounts (NRA) linked to China, which then sent large flows into UAE entities.

The report goes on to outline the involvement of a myriad of different types of companies integral to the network:

• Oil companies: entities trading in Iranian oil/petrochemicals used foreign front companies and USD correspondent banking to conduct business
• Shell companies: numerous indicators of shell-entity use (recent incorporation, shared addresses, no internet presence), which dominate the network’s structure
• Shipping companies: transacted ~$707 million (8% of total) for shipping-related services tied to Iranian oil transport, including ship-to‐ship transfers
• Investment companies: involved in routing funds (~$665 million, 7%) to give Iranian actors access to international investment markets
• Technology procurement companies: entities tied to purchasing export-controlled technology for Iranian military/nuclear programs received ~$413 million (5%)

UAE-, Hong Kong-, and Singapore-based companies are the backbone of the network

The United Arab Emirates (mainly the Emirate of Dubai) features dominantly in the flow of funds. About 71% (approximately $6.4 billion) of the total funds in the dataset were handled by UAE-based companies. Of that, UAE-based entities were receiving about $5.6 billion (62% of the total). Hong Kong-based shell companies (often using Chinese non-resident accounts) were the second largest origin point, sending around $4.8 billion (53%). Singapore accounted for 24% (~$2.2 billion) of total flows. The UK and Switzerland also appeared, although at far lower volumes (6% and 1% respectively).

U.S. correspondent banks exposed

Of particular note is the use of correspondent accounts at U.S.-based banks. FinCEN identified ~$534 million (6% of total) transferred from U.S. bank accounts to Iran-linked entities. Moreover, ~$361 million (4%) was sent via foreign branches of U.S. banks. These flows demonstrate that even U.S. banks must remain vigilant regarding indirect exposure to Iran-linked shadow networks and evasion schemes writ large.

From a compliance perspective, the report underscores several key red flags and areas of operational risk:

• Use of multiple layered shell companies and frequent wire transfers through jurisdictions with weak transparency (UAE free-zones, Hong Kong, Singapore).
• Transactions where the business purpose is vague or inconsistent with the corporate profile, often involving oil or shipping sectors.
• Significant activity involving U.S. correspondent banking channels, even if the ultimate beneficiary is Iran-linked.
• Prevalent use of free-trade zones or special purpose entities with opaque ownership to mask sanctioned-party exposure.

Using AI to counter sanctions evasion and disrupt hidden financial networks

With roughly $9 billion in suspicious flows linked to oil, shipping, technology procurement, and investment intermediaries, Iran’s shadow-banking typology underscores the scale and sophistication of sanctions evasion networks that traditional compliance systems – built around static rules, manual reviews, and list-based screening – are ill-equipped to detect. The report’s findings – particularly around UAE, Hong Kong, and Singapore intermediaries – highlight the urgent need for AI-driven detection, entity resolution, and behavioral analysis capabilities across the financial sector. AI helps institutions not only manage but get ahead of risks, ushering in a new era of proactive, intelligence-driven compliance.

AI-enhanced network analysis and entity resolution

Traditional screening systems rely heavily on static lists and name-matching, which often miss the layers of shell and front companies masking Iranian ownership or control. AI, particularly graph-based learning and natural-language processing (NLP), can dynamically map hidden relationships across entities, addresses, beneficial owners, and transactional patterns. By ingesting structured and unstructured data (e.g., corporate registries, trade documents, shipping manifests, and SAR narratives), AI can uncover indirect ties between seemingly unrelated entities operating across Dubai free zones or Hong Kong shell registries. Machine-learning-driven entity resolution can merge fragmented customer data into unified risk profiles, enabling banks to see the full network exposure rather than isolated accounts.

Behavioral analytics to identify shadow-network typologies

Machine learning models trained on past SAR data can learn the signatures of evasion activity: rapid fund layering through oil-trading intermediaries or weak jurisdictions, circular fund flows between UAE and East Asian entities, or funds routed through the same correspondent pathways for unrelated businesses. These models evolve as they ingest new data, providing a continuously improving radar for suspicious activity.

Risk prioritization and dynamic scenario optimization

AI can help compliance teams focus on what matters most. AI can score and prioritize alerts based on network centrality, transaction value, and proximity to known sanctions risks. This allows institutions to focus investigations where risk concentration is highest. Moreover, AI-driven scenario optimization can automatically tune thresholds in sanctions and AML-monitoring systems, minimizing false positives while maintaining regulatory coverage.

AI-assisted investigations and narrative generation

Generative AI can act as an analytical partner by summarizing SAR narratives, cross-referencing external data sources, and generating investigative summaries. These highlight linkages among counterparties, shipping routes, and payment chains. When applied responsibly, this accelerates analyst efficiency while ensuring consistent documentation for regulatory review. For U.S. correspondent banks in particular – where indirect exposure to Iran-linked funds poses compliance risk – AI-assisted triage and investigative tools can drastically improve response speed and quality.

Future-proofing sanctions programs through predictive intelligence

The next frontier for AI in sanctions and AML programs is prediction. By integrating geopolitical intelligence, trade data, and historical typologies, AI can forecast emerging evasion hotspots (such as new free-trade zones or sectors that may become proxies for sanctioned trade). This transforms compliance from reactive detection to forward-looking risk anticipation, a key evolution as sanctions networks evolve faster than regulatory updates and an expectation that regulators are increasingly signaling.

Real-time threat ID and detection recalibration

This FTA, along with many others issued by FinCEN and other global regulators, demonstrates the rich threat information that is available to help institutions first understand typologies of criminal behavior and then use that knowledge to recalibrate their risk management and monitoring approaches. AI can automate threat research, parse out risks, and recalibrate detections to ensure maximum coverage for a sound risk-based approach.

Traditional controls alone cannot keep pace with globally distributed AI-enabled adversaries. Financial institutions that harness AI for network mapping, behavioral analytics, and predictive modeling will be best positioned to detect complex evasion schemes, safeguard correspondent channels, and meet escalating regulatory expectations in an increasingly multipolar sanctions environment.

How Sensa Risk Intelligence reinvents compliance and risk management

SymphonyAI’s innovative technology reinvents how compliance and risk management work is completed. In this age of increasingly complex threats, which come with serious regulatory and reputational risks, AI and decades of industry experience is the answer. The Sensa Risk Intelligence (SRI) platform’s AI-native, modular design features agentic AI that intelligently automates and optimizes end-to-end processes. SRI’s Sensa Data, Sensa Detection, Sensa Agent, and Sensa Investigation is the strategic AI advantage that redefines your operating model and transforms compliance from a defensive function into a strategic asset.

Related resources

Introduction to Sensa Risk Intelligence

From reactive to proactive: Managing regulatory compliance with AI

Reinventing the compliance operating model

Command and control rewired: Agentic AI in anti-financial crime

SensaAI for Sanctions

References

[1] Financial Crimes Enforcement Network, “Financial Trend Analysis – Iranian Shadow Banking: Trends in Bank Secrecy Act Data,” October 2025