Financial institutions know collaborative fraud detection and shared intelligence are vital. But many still remain hesitant to share data across borders and competitors. The rules on what can and can’t be shared are often murky, and the liabilities tied to disclosure make collaboration feel risky.
Yet as regulators begin to clarify those boundaries, and privacy-preserving technologies make secure data collaboration possible, the question is changing from “Can we collaborate?” to “Can we afford not to?” Here’s why collaboration is key in modern fraud detection.
PETs are cryptographic and computational techniques, such as tokenisation, secure multi-party computation, and federated learning, that let organisations analyse or compare data without exposing its underlying contents. In fraud detection, PETs are what make it possible for banks to collaborate on risk signals without ever pooling raw customer data.
Effective collaboration gives financial institutions the broader visibility needed to identify fraud patterns that would otherwise remain hidden.
This is especially important when tackling money mule activity, where not all mules are created equal. Some orchestrate the fraud, while others are manipulated or deceived. Fraud is now a global, borderless enterprise that moves faster than individual institutions can respond. Criminal networks exploit the fragmented visibility between banks, fintechs, and payment providers, moving funds across multiple institutions and borders within seconds. While each organisation may only see a small part of the activity, a collaborative view reveals the full fraud pattern.
Modern fraud is also increasingly synthetic. Attackers use a blend of real and fabricated data to create identities that pass KYC checks and open accounts undetected. They link these synthetic profiles to webs of devices, IP addresses, and mule accounts, which makes rule-based monitoring far less effective. As a result, the industry’s prevention strategy can no longer rely solely on internal data or static controls.
Privacy-preserving data and intelligence-sharing arrangements allow banks to enrich their fraud detection models with digital footprint intelligence by correlating external indicators, such as device fingerprints, IP addresses, and email domains, with internal account and transaction signals - without exposing raw customer data.
A collaborative approach connects what each institution sees in isolation into a multi-dimensional risk picture. This helps teams distinguish genuine users from organized criminal activity much earlier in the process, strengthening overall fraud prevention.
Recent experiments by Swift showed how powerful a collective view can be. In trials involving ten million artificial transactions, a collaborative model trained across multiple institutions was twice as effective in detecting fraud in real time compared with models trained on data from a single bank.
Using privacy-enhancing technologies (PETs), participants securely shared artificial transaction data. PETs are cryptographic tools that allow data to be analysed or compared without revealing its underlying contents. Instead of exchanging raw transaction data, institutions share encrypted or synthetic representations that enable collaborative fraud detection while preserving customer privacy.
In a second use case, the team combined PETs with federated learning. This is an AI technique that trains algorithms locally within each institution, which allows models to learn from shared experience without ever exposing customer data.
Regulators are increasingly recognising the value of collaboration in strengthening fraud detection and preventing financial crime. In the UK, the Economic Crime and Corporate Transparency Act 2023 introduced new information-sharing measures that enable AML-regulated private-sector entities to share customer information with one another (or via intermediaries) for the purpose of detecting, preventing, or investigating economic crime.
Meanwhile, in the EU, the upcoming Anti-Money Laundering measures (effective July 2027), introduce a formal legal framework for firms to share financial crime intelligence through authorised ‘partnerships for information sharing’. By establishing consistent standards for cross-sector and cross-border data exchange and embedding data-protection safeguards, the regulation is expected to reduce the legal and operational uncertainty that has long restrained collaboration across institutions.
Collaboration offers significant opportunities to improve fraud detection, but it also introduces technical, operational, and regulatory challenges. Understanding both the benefits and the potential obstacles is essential for building secure and effective intelligence-sharing initiatives.
Even as regulators encourage information sharing, technical and operational barriers have kept large-scale collaboration rare. The industry’s first response has been to experiment with privacy-enhancing technologies (PETs) and generic privacy platforms designed to make data sharing safe. These tools show important progress because they prove that banks can collaborate without breaching confidentiality or regulatory boundaries.
Yet most of these platforms were never built for fraud detection. They were created to solve generic data-sharing challenges, for example, enabling research or marketing teams to analyze anonymized datasets, not to detect real-time financial crime. And while they protect privacy, they often fail to deliver actionable fraud intelligence or reflect how complex and fast-moving modern scams really are.
Fraud detection demands real-time, contextual, and highly specialized intelligence. Generic privacy platforms don’t understand the nuances of how fraud teams work — the evolving typologies, behavioral red flags, and real-world response requirements.
Every bank structures and labels its data differently, from account hierarchies to transaction types and device identifiers. In fraud detection, those differences matter. A privacy platform might anonymize the data but still fail to align the meaning behind it.
Traditional “consortia” models often amount to little more than shared blacklists of flagged accounts or entities that quickly become outdated. Meanwhile, scammers adapt in real time, creating new synthetic identities and building money mule networks that easily bypass these static defences. Because these systems depend so heavily on the skill and availability of each institution’s data scientists, collaboration remains fragmented and inconsistent.
Generic privacy platforms protect data but rarely stop fraud. They anonymize and transport information safely, yet they don’t understand the context or velocity of criminal behaviour. Acoru closes that gap by combining privacy-preserving architecture with the fraud-detection expertise and real-time orchestration that those platforms lack.
Instead of pooling sensitive customer data into a shared database, Acoru’s Consortium Manager enables participating institutions to consult one another in real time securely, selectively, and under each bank’s own policy controls. Each request stays privacy-first: no personally identifiable information (PII) is ever pooled or exposed.
Each response is explainable, signed, and auditable, so banks can act confidently, knowing they have verifiable intelligence that meets regulatory standards. Anonymized or tokenized account and behavioural signals are exchanged under strict policy enforcement, meaning collaboration doesn’t come at the expense of compliance.
Where generic privacy tools stop at anonymization, Acoru builds in fraud-specific intelligence:
Beyond data sharing, Acoru’s Consortium Manager also enables federated model training, allowing participants’ fraud detection models to improve through shared learning without compromising privacy.
Effective collaboration requires more than secure data sharing. Financial institutions need solutions that combine privacy-preserving collaboration with fraud-specific intelligence, enabling them to detect threats earlier while maintaining privacy, compliance, and operational control.
Discover how Acoru's Consortium Manager helps financial institutions collaborate securely to detect fraud earlier.
Collaborative fraud detection is the practice of financial institutions sharing fraud signals, risk intelligence, or model training data with one another, securely and often via privacy-preserving technology, to identify fraud patterns that no single institution could see using only its own data.
PETs are cryptographic and computational tools, such as tokenisation and federated learning, that let organisations analyse or compare data without exposing the underlying contents, making it possible for banks to collaborate on fraud signals without pooling raw customer data.
Increasingly, yes, within specific frameworks. In the UK, the Economic Crime and Corporate Transparency Act 2023 created new information-sharing gateways for AML-regulated entities. In the EU, Article 75 of the AML Regulation introduces "partnerships for information sharing" from July 2027. In the US, Section 314(b) of the USA PATRIOT Act has provided a safe harbour for this kind of sharing since 2001.
Most were built for research or marketing use cases, anonymising data for analysis, not for real-time financial crime. They typically don't understand fraud-specific context (typologies, behavioural red flags, how account and transaction data differ from bank to bank), so they protect privacy without necessarily producing actionable fraud intelligence.
No. Modern approaches, including federated learning and tokenised or signed data exchange, allow institutions to benefit from shared intelligence and shared model training without personally identifiable information ever leaving the originating institution.