In fact, 380 million people use BNPL services globally. This is predicted to increase to 670 million by 2028. This rapid growth brings fresh challenges: more users, higher transaction volumes, and lighter onboarding controls create fertile ground for fraudsters to test and exploit gaps in risk controls. 

For payments professionals, that means balancing the demand for fast, frictionless customer experiences with the need to protect revenue, manage chargebacks, and reduce operational strain.

Let’s break down what BNPL fraud looks like in practice, explore why it’s a growing concern for ecommerce teams, and share practical strategies your business can adopt to prevent fraud without undermining conversion.

What Is BNPL Fraud and Why It’s a Growing Concern

BNPL fraud refers to activity that exploits Buy Now, Pay Later payment systems to deceive merchants or providers and obtain goods, services, or credit without intending to pay back. Fraudsters take advantage of BNPL’s lightweight onboarding and delayed settlement structure, making it easier to abuse the system compared with traditional payment methods.⁴⁾ Common examples include:

• Identity misuse and synthetic identities — criminals use stolen personal information or fabricated identities to open BNPL accounts or complete purchases.
• Friendly or first-party fraud — an initially legitimate user who abuses the system by taking out BNPL credit with no intention of repaying or disputes legitimate transactions to force a refund.
• Account takeover (ATO) — fraudsters compromise a legitimate user’s account to make purchases without the owner’s knowledge.

Similarly to traditional card fraud, BNPL fraud often begins in the onboarding process and unfolds over weeks or months, because the payment is deferred and the provider only learns of non-repayment later. This difference makes detection and response more complicated than instant card-based transactions.

Why BNPL Fraud Is Growing

Several structural features of BNPL heighten the fraud risk:

• Frictionless onboarding lowers barriers: BNPL services prioritise speed and ease at checkout, often with limited identity checks, which fraudsters can exploit to slip through risk controls.

• Rapid volume growth expands attack surface: As BNPL adoption continues to surge worldwide, fraudsters increasingly target these platforms with both traditional tactics and new schemes like bot-based application attacks.-making according to Visa.  

The combination of convenience for legitimate customers and exploitable vulnerabilities for bad actors is why BNPL fraud continues to rise, making robust prevention strategies essential for payments and fraud teams. 

Top BNPL Fraud Problems and Why Traditional Controls Struggle

BNPL fraud continues to grow as Buy Now, Pay Later becomes a more common payment method in ecommerce. The same features that make BNPL attractive to shoppers — fast onboarding and deferred payments — also introduce gaps that fraudsters can exploit. Traditional fraud controls, designed for older payment models, often aren’t built to address these unique dynamics. 

Identity Abuse and Synthetic Identities

One of the most common BNPL fraud problems is identity exploitation. Fraudsters use stolen personal information or create synthetic identities by combining real data with fake details. 

Because many BNPL systems aim for minimal friction at signup, these fraudulent profiles can slip through basic checks, be used to make purchases, and then be abandoned before repayment is due.

Strategic Misuse by Customers

Not all BNPL fraud originates outside the customer base. Friendly fraud occurs when a shopper disputes a legitimate BNPL transaction, claiming it was unauthorized in order to trigger a refund. 

Because the customer is legitimate and uses legitimate payment methods, resolving these cases can be trickier than straightforward fraud cases, and can increase operational workload and losses for merchants.

Account Takeover and Rapid-Fire Attacks

Another frequent risk is account takeover (ATO), where fraudsters gain access to legitimate user accounts through stolen credentials, phishing, or credential stuffing. Relatedly, automated attacks can submit multiple BNPL applications in quick succession from the same device or network in an attempt to overwhelm simple rule-based systems. 

Traditional fraud tools, optimised for instant credit card authorisation, often miss these patterns because the risk doesn’t fully manifest at the moment of checkout.

Why Traditional Controls Fail to Tackle BNPL Fraud

Legacy fraud systems primarily rely on static rules and historical payment signals that work reasonably well for immediate card transactions. They struggle when signals are subtle, delayed, or behavioural in nature, common characteristics of BNPL fraud. 

They also often force a trade-off between false positives (blocking legitimate customers) and false negatives (missing real fraud), which can hurt conversion rates and customer experience.

This combination of new fraud patterns and outdated controls is why many teams are moving toward modern, adaptive BNPL fraud prevention tools that can assess risk in real time, incorporate identity intelligence, and monitor behaviour throughout the payment lifecycle.

How to Improve BNPL Fraud Prevention

Effective BNPL fraud prevention requires more than adapting traditional card-based controls. Because Buy Now, Pay Later models combine fast onboarding with deferred repayment, risk teams need layered, real-time approaches that assess identity, behaviour, and transaction context together.

Strengthen Identity Verification at Onboarding

Lightweight signup processes are attractive to customers and fraudsters alike. That’s why identity verification should go beyond basic email and address checks. This includes evaluating device intelligence, email and phone reputation, behavioural signals and consistency across data points. 

The goal is not to introduce heavy friction for every customer, but to build a clearer picture of whether a user is genuinely who they claim to be.

Use Real-Time Risk Scoring at Key Decision Points

BNPL fraud often begins at account creation or first purchase. Real-time fraud detection at onboarding and checkout allows teams to assess risk instantly, rather than waiting for repayment failures or disputes to reveal problems later. 

Adaptive risk scoring, based on identity signals, transaction context, and behavioural patterns, helps teams act before losses occur.

Monitor Velocity and Anomalous Behaviour

Fraudsters frequently test systems by submitting multiple applications, using the same device across different identities, or making rapid, high-value purchases. 

Effective BNPL fraud prevention includes monitoring for velocity spikes, unusual account patterns, and inconsistencies across sessions. These behavioural indicators often reveal organised activity that static rule sets miss.

Extend Controls Beyond the First Transaction

Because BNPL revenue is deferred, fraud risk doesn’t end at checkout. Ongoing monitoring — including changes in device, payment details, delivery address, or repayment behaviour — helps identify account takeover, friendly fraud, or emerging repayment risk before it escalates. Continuous fraud detection reduces the likelihood that issues only surface weeks later.

Balance Risk Reduction with Customer Experience

Perhaps the most important principle is balance. Overly aggressive controls can create high false positives, blocking legitimate buyers and damaging conversion rates. 

Intelligent BNPL fraud prevention uses risk-based step-ups, applying stronger checks only when signals justify them, to protect revenue without undermining the seamless experience customers expect.

Why This Matters for Payments and Fraud Teams

Stronger BNPL fraud prevention delivers more than lower fraud rates;

It directly reduces losses, chargebacks and write-offs, protecting margins in a payment model where repayment risk is already deferred.

Secondly, better fraud detection lowers manual review volumes, reducing operational strain and allowing teams to focus on higher-value investigations rather than processing borderline cases.

It’s also worth remembering that by minimising unnecessary declines, you improve approval rates and customer experience, supporting conversion and repeat purchase behaviour.

Finally, more sophisticated risk controls strengthen your compliance and partner relationships, demonstrating responsible credit management in a regulatory environment that continues to evolve.

In short, effective BNPL fraud prevention is not just a defensive measure — it’s a strategic enabler of sustainable growth. By reducing losses while preserving customer experience, payments teams can support long-term revenue expansion without increasing exposure to fraud risk.

Start Building your BNPL Fraud Protection Strategy

The answer isn’t heavier, more disruptive checks. It’s balanced, intelligent BNPL fraud prevention that combines strong identity signals, real-time fraud detection and ongoing monitoring without undermining the customer experience.

If done well, these measures shouldn’t affect your growth, they should actually complement it. As BNPL adoption continues to expand, teams that modernise their approach to fraud risk will be best positioned to reduce losses, maintain approval rates and support sustainable revenue growth.

For deeper insight into related identity and payment fraud challenges, explore our other resources on digital identity verification and fraud prevention strategies.

If I (Sébastien Carletti) were to elaborate on the BNPL fraud subject, here would be my take:

BNPL Fraud Prevention: Key Risks and How Businesses Can Reduce Exposure

What is BNPL Fraud?

Buy Now, Pay Later (BNPL) solutions have rapidly transformed the ecommerce landscape, allowing customers to split payments into instalments while merchants benefit from higher conversion rates and larger basket sizes.

However, the rapid growth of BNPL has also attracted new forms of fraud and misuse. The deferred nature of the payment creates a structural gap between the purchase and the moment the financial risk materialises.

Traditionally, fraud in digital payments refers to situations where an attacker impersonates or takes control of a legitimate user in order to obtain goods or services without paying. In BNPL environments, this remains a major threat.

But BNPL also introduces a second category of risk: situations where the customer is genuine and correctly identified, yet repayment may never occur.

Understanding BNPL risk therefore requires distinguishing between professional fraud and strategic misuse by legitimate customers.

Why BNPL Fraud is Increasing

Several factors contribute to the growing exposure of BNPL providers.

Frictionless onboarding

BNPL services are designed to reduce friction at checkout. While this improves customer experience, it also lowers the barrier for attackers attempting to create accounts or apply for credit using stolen or synthetic identities.

Deferred payment models

Unlike traditional card payments, the financial risk in BNPL appears later in the transaction lifecycle. This delay gives both fraudsters and opportunistic users more time before repayment obligations become visible.

Rapid market adoption

As BNPL becomes a mainstream payment option, the attack surface naturally increases. More users, more merchants, and more transactions mean a broader environment where both organised fraud and behavioural misuse can occur.

Key BNPL Risk Categories

The post BNPL Fraud Prevention: Key Risks and Solutions  appeared first on Oneytrust.

So, unsurprisingly, fraud in digital ecosystems has evolved to overcome existing checks and safeguards.

Instead of breaking systems, criminals now move through them – using valid data, real registration processes and increasingly sophisticated methods to appear legitimate.

Common tactics include:

• Creating shell or short-lived companies designed to pass verification (including KYB)
• Impersonating legal representatives or directors of legitimate businesses
• Submitting real or near-authentic documentation
• Combining accurate data with false or inconsistent identity elements
• Exploiting fast onboarding processes to avoid deeper scrutiny

Such profiles are not obviously fraudulent. Every element is designed to blend in. And that is exactly why they succeed.

This is where subtle inconsistencies – otherwise known as ‘weak signals’ – can reveal what traditional checks overlook.

The structural limitations of KYB
If you’re in B2B and regularly onboarding new partners/merchants, you’re already aware of Know Your Business (KYB). KYB’s mandatory compliance processes were designed to answer one question: does this company exist?

But existence is not the same as legitimacy. It doesn’t mean a company is safe to work with. This is especially evident in cases of synthetic identity, where real and fake data are combined to create profiles that pass standard verification unnoticed. You can read more on synthetic identities here.

KYB verifies registration data, legal documents and company status. But it does not verify:

• Whether the person acting on behalf of the company is legitimate
• Whether the identity and the business are actually connected
• Whether the digital behaviour behind the application is consistent

In other words, KYB validates the company on paper – but not the reality behind it.

This creates a blind spot that fraudsters thrive in.

Where modern fraud operates
In most B2B fraud scenarios, the issue isn’t the company alone. It’s the gap between three key elements:

• The company
• The representative(s)
• The digital footprint

Each may appear valid in isolation. But together, they don’t always align. For example:

• A legitimate company… paired with an impersonated director
• A real registration… linked to a disposable email and unusual device data
• A consistent identity… but behaviour that doesn’t match a real business user

Individually, these signals may not trigger concern. But combined, they reveal serious risk.

This lack of coherence is what traditional approaches fail to detect.

The grave cost of getting it wrong
We don’t need to hammer this home too much, but when these inconsistencies go unnoticed, the consequences are often significant:

• Fraudulent companies gain access to platforms, services, marketplaces, etc.
• Credit may be issued to entities that will never repay
• Equipment or goods may be rented and never returned
• Merchant accounts can be used for illicit activity
• Your business faces increased exposure to AML and compliance risks

Beyond financial loss, there is also reputational damage and operational disruption to worry about. These are often only discovered after the fraud has occurred.

A shift in how corporate identity is verified
To address evolving fraud techniques, businesses need to move beyond the document-based verification and other limitations inherent to KYB.

The question is no longer: “is this company valid?”, but rather: “does everything about this business – and the person behind it – make sense together?”

A longer question, sure, but an incredibly important one.

This requires a new approach to corporate identity verification, built on three principles:

1/ Verifying the company
Not just its existence, its sector of activity, whether it is facing insolvency proceedings, or its legitimacy.

2/ Verifying the representative
Confirming the identity of the individual(s) acting on behalf of the business

3/ Analysing digital behaviour
Understanding the signals behind the interaction, including device, IP, email and behavioural patterns

The real value comes from connecting these elements – and flagging where they don’t align.

Coherence is key
Today, digital identities aren’t just a set of static data – but a combination of attributes, behaviours and interactions that form a picture over time.

Modern fraud detection is no longer about checking isolated data points. It’s about assessing coherence. In other words… does the company match the person? Does the person match the behaviour? Does the overall profile resemble a legitimate business interaction?

This shift from validation to consistency allows businesses to detect fraud that would otherwise go unnoticed.

Closing the gaps in B2B onboarding
As onboarding becomes faster and more automated, the risks increase.

Manual checks can’t scale, document verification alone isn’t enough and fraudsters adapt faster than traditional controls.

To stay ahead, businesses need to rethink how they verify professional identities – bringing together company data, identity verification and digital signals into a single, coherent view.

This is where a new approach to corporate identity verification becomes essential.

A new approach to corporate identity verification
D-Risk ID Corporate is designed to address all the challenges we’ve discussed in this blog. In real-time, it verifies the company, the person behind it and the countless digital signals that underpin each application.

By closing the gaps left by traditional KYB, D-Risk ID Corporate enables businesses to:

• Detect fraud earlier in the onboarding process
• Reduce exposure to financial and compliance risk
• Approve legitimate businesses faster
• Build trust in high-volume business environments

Discover how D-Risk ID Corporate works here.

Conclusion
Businesses can’t beat B2B fraud by focusing on fake data alone.

The future of fraud prevention lies in understanding not just what a business claims to be – but whether everything around it makes sense.

So, while KYB verification focuses only on the company, D-Risk ID Corporate uncovers inconsistent identities – even those that appear legitimate.

The post Why KYB is no longer enough to stop B2B fraud appeared first on Oneytrust.

These figures illustrate a major shift: identity fraud is becoming an industrialized phenomenon, fueled by AI and organized on a large scale.

In this context, the central question for financial institutions and digital platforms is no longer simply how to strengthen controls, but how to detect fraudulent behavior in an environment where identity artifacts are becoming increasingly credible.

The pitfall of visible checks

Historically, the fight against fraud has been structured around visible checkpoints:

• ID checks
• facial recognition
• KYC procedures during onboarding

These mechanisms remain essential. But they share a fundamental characteristic: they are visible to those undergoing them. In other words, fraudsters can analyse them, test them and gradually learn to circumvent them. A fraudster regularly confronted with a face-matching system will quickly seek to understand how it works. They can then experiment with different methods: deepfakes, face swapping, video injection or biometric replication.

This phenomenon is nothing new. In many areas of security, visible mechanisms are always eventually analysed and circumvented. It is precisely for this reason that the fight against fraud cannot rely solely on what might be called the ‘right hand’: visible monitoring.

The ‘left hand’: invisible detection

An effective anti-fraud strategy actually relies on two complementary aspects:

• The ‘right hand’: verification. These are the visible checks that enable information or an identity to be confirmed.
• The ‘left hand’: detection. This involves analysing signals that are invisible or difficult for fraudsters to interpret.

The difference is crucial. Controls can be observed and circumvented. Detection relies on data patterns and correlations that are difficult to anticipate. In this model, controls play an important but secondary role: they serve to reinforce or confirm a signal detected elsewhere. This approach helps to avoid a constant technological arms race between fraudsters and control systems.

Deepfakes: the new frontier in biometric fraud

The emergence of deepfakes perfectly illustrates this trend. Modern biometric systems now incorporate liveness detection mechanisms, requiring the user to perform an action in real time. These checks make fraud more complex, but they also drive attackers towards more advanced techniques. According to Entrust, deepfakes now account for around 40% of fraud attempts on video biometric systems.

Fraudsters exploit, in particular:

• AI-generated deepfakes
• injection attacks using virtual cameras
• manipulated video streams inserted into capture systems

These techniques enable falsified biometric data to be fed directly into digital identity systems. Fraud is therefore no longer simply a matter of forging a document. It involves manipulating the data streams themselves.

The rise of synthetic identities

At the same time, AI is accelerating the creation of synthetic identities. Rather than stealing an existing identity, fraudsters create a new identity by combining:

• real personal data
• fabricated information
• manipulated or generated documents

AI tools enable these identities to be produced quickly, cheaply and on a large scale. This type of fraud is particularly dangerous because these identities can remain active within an organisation’s systems for a long time before being exploited. Losses associated with synthetic identity fraud are expected to reach tens of billions of dollars in the coming years.

AI for detection

In light of these developments, the issue is not simply a matter of using AI to strengthen visible controls. A more effective strategy involves deploying AI at the heart of detection mechanisms. Fraud does not usually manifest itself as a single anomaly. Rather, it emerges through an accumulation of weak signals:

• inconsistencies in data
• atypical behaviour
• correlations between accounts or identities
• similar activity patterns

Analysing these signals requires processing large amounts of data and detecting patterns invisible to the human eye. It is precisely in this area that artificial intelligence can deliver the greatest value. This approach fits within the DIKW (Data – Information – Knowledge – Wisdom) conceptual framework, where data analysis gradually transforms raw data into actionable knowledge.

A multi-layered anti-fraud architecture

To be effective, a modern anti-fraud strategy must combine several layers of protection:

• document analysis
• facial biometrics
• device and behavioural intelligence
• geolocation and velocity analysis
• identity correlation
• repeat fraud detection

The aim is not to increase the number of visible checks, but to combine multiple signals that make fraud detectable without being easily observable. This approach helps maintain a vital balance: protecting organisations against fraud whilst minimising friction for legitimate users.

Building trust in digital identity

In an increasingly digital world, identity verification remains a critical stage in the user journey. Onboarding often presents the first opportunity to build trust in an identity. Modern digital identity verification solutions now combine multiple sources of information and detection mechanisms to assess the overall risk associated with an identity.

Platforms such as Oneytrust’s D-Risk Commerce and D-Risk ID follow this approach by orchestrating various risk signals to enhance fraud detection without compromising the user experience. The aim is not to replace controls, but to integrate them into a broader detection strategy that is less predictable for fraudsters.

Conclusion

The rise of AI-generated deepfakes marks a new stage in the evolution of identity fraud. Fraud is no longer just a matter of forged documents or stolen data. It is becoming a systemic phenomenon combining synthetic identities, AI-generated content and automated attack infrastructures. In this context, organisations must move beyond an approach focused solely on visible controls.

True effectiveness lies in the ability to detect fraudulent patterns through data analysis and the identification of weak signals.

In other words, fighting fraud is not waged solely with the right hand — that of control. It is waged above all with the left hand — that of detection.

The post The Rise of AI-Generated Deepfakes in Identity Theft appeared first on Oneytrust.

Fraud is evolving rapidly thanks to generative AI

Fraudsters now have access to technologies that were once the preserve of experts. It takes just a few clicks to generate:

• doctored images of products “supposedly” broken,
• forged supporting documents,
• and well-crafted synthetic identities

As TF1 explains, private sellers are now facing disputes based on AI-manipulated photos, with automated systems sometimes turning against the honest victim themselves.

One user interviewed recounted losing €180 due to an AI-generated image showing a supposedly broken vinyl record — an item that, by its very nature, does not break in that way. The platform’s automated system, however, ruled in favour of the fraudster, illustrating the vulnerability of traditional checks.

This phenomenon is part of a global trend: fraud involving doctored images is skyrocketing, with a 15% increase in falsified images in claims since the start of 2025.

Three major trends observed by Oneytrust

At Oneytrust, our experts have identified three major shifts:

1. More sophisticated fraud

Generative AI can produce highly realistic images, some of which are impossible to distinguish with the naked eye. Synthetic identities or invoices recreated pixel by pixel can fool traditional verification systems.

2. The industrialisation of attempts

What was once manual work is now automated. Fraudsters are now launching vast, coordinated campaigns, drastically increasing the volume of attempts detected on e-commerce sites.
This industrialisation is confirmed by the TF1 report, which warns of waves of scams potentially affecting millions of people.

3. New weak signals

Traditional methods are no longer sufficient. When evidence is generated by AI, behaviour, metadata and contextual inconsistencies become the new areas of analysis.

 AI: an essential tool for countering these new attacks

With fraudsters now using AI, anti-fraud solutions must evolve. That is why Oneytrust has integrated advanced artificial intelligence models at the heart of its technologies.

In particular, our systems enable us to:

Detect behavioural and contextual inconsistencies

Frequency of returns, unusual histories, multiple addresses or devices: the patterns speak for themselves.
Our solution cross-references these signals to identify abuse well before it becomes visibly repetitive.

Adapt in real time

Fraud methods change every week. Our models continuously recalibrate to keep pace with these changes and block emerging attacks.

Identify organised networks

By analysing data structure and correlating identity details, devices and network history, Oneytrust exposes organised criminal groups operating under different identities or accounts.

Tailor detection to your returns processes

Oneytrust adapts precisely to each retailer’s returns policies and internal workflows, incorporating their rules, thresholds and operational specifics.

The real challenge: maintaining trust

In this context, the issue is no longer simply about blocking fraud.
It is about protecting the relationship of trust between retailers and their customers.

Excessive suspicion can damage the shopping experience. Being too permissive can encourage abuse and undermine business models. The balance is therefore delicate — and AI plays a crucial role in accurately distinguishing between:

• honest shoppers,
• opportunistic fraudsters,
• and organised criminal networks.

Oneytrust is committed to maintaining this balance by combining cutting-edge technology with human expertise.

Fraud and AI: a huge challenge… but far from insurmountable

Artificial intelligence has profoundly transformed the landscape of reimbursement fraud. Whilst it provides fraudsters with new tools, it further enhances companies’ ability to detect and stop them.

With ongoing investment, increasingly sophisticated models and greater collaboration across the entire ecosystem, this battle is not only possible, but winnable.

Trust is one of the cornerstones of online commerce. Together, let’s continue to protect it.

The post Refund fraud: when AI becomes a tool for fraudsters… appeared first on Oneytrust.

1. The new broad definition of an ‘AI system’

The text defines an AI system very broadly as a ‘machine-based system, operating with varying degrees of autonomy, capable of adapting after deployment and which, for explicit or implicit purposes, deduces from data how to generate results (predictions, content, recommendations, decisions) that influence physical or virtual environments’.

This broad technological scope is complex to implement operationally, but the market agrees that it encompasses sophisticated approaches such as machine learning (ML), hybrid expert rule + machine learning approaches, and advanced optimisation.

With the proliferation of AI use by fraudsters, trusted anti-fraud companies such as Oneytrust will need to further develop their range of AI solutions to power their scoring, behavioural alerting and anomaly analysis engines in order to keep up with the times and identify increasingly sophisticated fraud patterns.

2. Risks: where does fraud detection fit in?

The AI Act classifies certain uses as ‘high risk’ (critical biometrics, employment, education, access to essential services such as credit scoring, health insurance pricing, etc.). These systems are authorised but must be accompanied by a very sophisticated risk management system and an assessment of their compliance before being placed on the market.

One notable exception is that AI used to detect financial fraud is not automatically considered ‘high risk’. This reduces the direct regulatory burden, but does not remove the expectations of transparency, data quality and human oversight that regulated institutions contractually pass on to their subsidiaries and suppliers (such as Oneytrust).

3. What are the key dates to remember?

The main requirements of the AI Act will come into force in stages. Here are the key dates to remember:

• 2 February 2025: entry into force of the ‘unacceptable risk’ prohibitions + AI literacy requirement (awareness/training for staff involved in AI).
• 2 August 2025: obligations for general-purpose AI models (GPAI) begin to apply (generative AI).
• 2 February 2026: European deadline for certain implementing acts (post-market surveillance plans).
• 2 August 2026: obligations for high-risk AI systems.
• August 2027: general application of all provisions of the regulation.

4. Why prepare even if you are not ‘high risk’?

Fraudsters are becoming industrialised with AI. Deepfakes, fake documents, automated attack scripts, automated use of synthetic identities: fraud is scaling up. Those involved in the fight against fraud must adapt quickly to these new trends.

Although fraud prevention is not explicitly considered high risk, it is essential to ensure that solutions respect the fundamental rights and privacy of customers and end users, whether in accordance with the GDPR or the AI Act. In addition, the expectations of regulated partners (mainly in the banking sector) are increasingly high, as they are subject to sector-specific requirements that demand traceability, good documentation and heightened vigilance regarding the quality of the data used to satisfy their supervisors.

5. The Oneytrust response

Oneytrust provides identity verification and fraud detection solutions for e-merchants, fintechs and banks, relying on AI solutions coupled with our 25 years of human expertise in fraud prevention to detect synthetic identities, transactional anomalies and risk signals in real time. We are members of the BPCE Group, which drives us to high standards of compliance and model governance.¹

As with the GDPR, we did not wait to cultivate our expertise on the AI Act and have been raising awareness and training our staff on these new compliance issues for several years. Contact us if you would like to learn more about Oneytrust’s vision for AI compliance and risk management!

The post AI Act and Fraud Prevention : why the market must prepare appeared first on Oneytrust.

That something is weak signals.

What is a weak signal in digital identity?

A weak signal, as its name suggests, is not a red alert. It is a detail. A micro-anomaly. An almost invisible (or rather ignored) inconsistency taken in isolation, but which takes on its full meaning when correlated with other elements.

For example:

• a known device… but whose technical fingerprint has changed slightly;
• a connection from a usual city… at a completely atypical time;
• correct keystrokes… but with an unusual rhythm;
• a smooth user journey… but traveled at a speed too fast to be human.

None of these elements alone is sufficient to justify blocking an event. But together, they tell a different story from what the declarative data would suggest.

Why strong signals are no longer enough

Historically, security relied on binary elements: correct password, validated SMS code, recognized device.

The problem? These proofs can be stolen, intercepted, or simulated.

Malware, phishing, massive data leaks, and automation tools have turned credentials into mere raw material for fraudsters. Even strong authentication is no longer foolproof against real-time proxy attacks or remote takeover.

The result: a user can tick all the boxes for legitimacy while still being a fraudster. This is precisely where weak signals become decisive.

Weak signals reveal behavior, not a declared identity

A declared identity can be falsified. Behavior is much more difficult to fake.

Weak signals allow us to answer not the question “Is this information correct?” but “Does this behavior resemble that of a legitimate human being in this specific context?”

This leads us to a dynamic interpretation of digital identity, based on the consistency of habits, the continuity of interactions, and the logic of sequences of actions.

From isolated events to behavioral history

The true power of weak signals lies not in a snapshot, but in duration.

A strange connection may be insignificant. Two anomalies close together begin to raise questions. A series of micro-deviations paints a picture of fraud.

It is by connecting these dots that we move from one-off checks to continuous behavioral monitoring. Digital identity ceases to be a state and becomes an evolving probability.

The challenge: detecting without degrading the experience

Exploiting weak signals does not mean increasing friction.

The goal is to adapt the level of vigilance: request additional verification only when the risk increases, silently monitor suspicious behavior, and trigger enhanced controls on sensitive actions.

In other words: make security proportional to the actual risk.

When email addresses become a weak signal

Email addresses are often perceived as simple contact identifiers. However, they contain a wealth of information that is underestimated when analyzed from a semantic and behavioral perspective.

Let’s take two examples: marie.dupont@yahoo.fr and ma.rie_du.pont1567@gmail.com.

Both addresses are technically valid. Both can pass standard checks. But they don’t tell the same story.

Semantic analysis involves observing the structure, logic, and consistency of an email address in relation to the context declared by the user.

Discreet but telling clues may emerge: domain name, username composition, presence of promotional keywords, consistent or inconsistent series of numbers, unusual complexity, or recurrence of similar patterns observed in previous fraud cases.

Taken in isolation, none of these elements proves fraud. But they may indicate that an address was created quickly, en masse, or for purely transactional rather than relational purposes.

A real digital identity often leaves traces of continuity. Conversely, in many fraud scenarios, the email address is a disposable tool designed to pass a technical check, not to reflect a lasting identity.

Semantic analysis therefore allows us to answer a subtle question: does this address resemble a human point of contact or a functional artifact created for a specific scenario?

Once again, the value comes from correlation: unusually structured email, rapid account creation, device recently seen on other accounts, automated browsing. It is the aggregation of these weak signals that reveals a credible risk.

From raw data to contextual intelligence

Weak signals already exist in your systems: technical logs, browsing data, device metadata, action sequences.

The difference lies not in the quantity of data, but in the ability to correlate it in real time, contextualize it according to the journey, learn from past patterns, and detect subtle deviations.

Digital identity is no longer a file, it is a flow

For years, we treated identity as a file: static information to be verified once and for all.

Weak signals are forcing us to change our paradigm. Identity is becoming a continuous stream of behaviors, which are confirmed or degraded through interactions.

Conclusion

Major frauds rarely hide behind big anomalies. They are concealed in the details that we don’t look at.

Learning to read weak signals means accepting that the truth is no longer found in a single piece of evidence, but in the accumulation of micro-clues. In a world where identities can be fabricated, stolen, or rented, this subtle reading becomes one of the pillars of digital trust.

The post Weak signals: those little-noticed details that reveal the authenticity of a digital identity appeared first on Oneytrust.

Across Europe according the MRC online merchants are already losing roughly 2.8% of revenue to fraud, with fraud representing about 3% of all orders. Identity fraud and account takeover are surging, fuelled in part by AI that can generate convincing fake identities and social-engineering scripts at scale according to UK fraud agency CIFAS. Agentic AI simply supercharges that trend on both sides of the fence. 

For merchants and finance companies though – there is a dilemma. Increasingly buyers and new customers will use AI to seek out and purchase goods and services,  

From now you’ve basically got two new “customer segments” turning up in your data: 

• Legitimate agentic AI – shopping assistants, payment agents and aggregators acting on behalf of real people. 
• Malicious agentic AI – computer-using agents hammering your sign-up flows, ID checks and checkouts at scale. 

The job for merchants and financial services providers is to work with their fraud detection companies to pick their way through this minefield. At Oneytrust we have been seeing this emerge over 2025 and like the rest of the world we anticipate a huge uptick in 2026.  

So, what you do not want to block; agentic AI is already being used as a consumer interface for payments and shopping: 

• A number of agentic AI options are emerging in Europe like Mastercard “Agent Pay”, Visa “Intelligent Commerce”, Amazon “Buy for Me” and Google “Shop with AI”, where agents initiate payments within parameters set by the consumer 

• Payments players describe “agentic commerce” as AI agents that hold conditional permissions to shop and pay on a user’s behalf, functionally similar to cards-on-file or recurring payments but with a conversational UX and more decision-making according to Visa.  

Regulators are behind the curve: PSD2 doesn’t mention AI, and even PSD3/PSR only references AI once for fraud prevention. But consumer agents are happening anyway. 

For a merchant, that means some “bot traffic” is now high-value, compliant traffic (AI doing what a loyal customer asked it to do). These agents will often come from data-centre IPs, headless browsers or shared devices, and will reuse stored credentials and payment instruments – i.e. very similar to the bots you’ve spent so much time trying to deter.  

If you block everything that looks like an agent, you’ll break this emerging channel and annoy the schemes, PSPs and big-tech partners driving it. 

How fraudsters are using agentic AI against merchants and banks 

Several trends are showing up in European and global reporting: 

Credential stuffing & ATO with Computer-Using Agents (CUAs). 

According to a report by Push Security,  OpenAI “Operator”-style CUAs shows they can log in to arbitrary web apps, read pages, click buttons and handle full flows like a human – but at bot scale. That lets attackers spray stolen credentials across thousands of sites and then perform in-app actions once they get in. 

AI-scaled phishing and social engineering feeding into payments. 

The European Payments Council’s 2025 threats report flags AI-generated phishing and deepfakes as a key enabler of APP fraud and impersonation scams, making language barriers vanish. 

Account opening and synthetic ID at industrial scale. 

Financial-crime specialists warn that agentic AI will be used to flood banks’ online onboarding with highly consistent, multi-step new account applications, reusing and recombining stolen or synthetic identity elements. 

Payment fraud “speedruns”. 

Arkose Labs talk about agents that skip normal browsing, go straight to high-value endpoints (card testing, gift cards, BNPL, high-ticket items), and adapt in real time if they hit friction. 

Your ID and fraud stack is going to see AI agents trying to open accounts and make purchases 24/7, some benign, some absolutely not. 

Why classic bot defences aren’t enough 

The nasty twist, well-summed up by one recent fraud blog, is that beneficial and malicious agentic AI are technically indistinguishable at first glance. Both: 

• Run in browsers or CUAs that move the mouse, scroll and click like humans. 
• Can introduce jitter into timings and keystrokes. 
• Can respect (or deliberately emulate) your UX flows. 

Old-school bot rules – “data-centre IP = block”, “too fast = bot”, “no mouse = bot” – are blunt instruments here and will kill legitimate payment agents along with attackers. You have to stop thinking “bot vs human” and start thinking in terms of intent and pattern at the identity, device and journey level 

How to tell good agents from bad ones 

For merchants and banks using an ID+fraud provider like Oneytrust, the detection strategy for this use case looks roughly like this. 

Treat “agent” as its own identity class 

First step is to explicitly model agent traffic: 

Tag sessions where the user agent, device behaviour or integration pattern clearly indicates an AI or automation layer. 

Maintain separate risk baselines for: 

Human sessions 

First-party agents (your own app’s automation) 

Trusted third-party agents (big-tech payments, official partners and the new AI purchase agents listed above) 

Unknown / suspicious agents 

Legitimate agents tend to be stable over time – same provider, similar IP ranges, same small set of identities, and predictable timing – the same as legitimate people. Malicious agents show sprawl across identities, merchants and institutions – mimicking their fraudster creators.  

Identity and data-level coherence 

This is where your digital-identity layer comes into its own, Oneytrust’s own positioning is crystal clear: as generative AI, deepfakes and synthetic ID fraud rise, static KYC checks are turning into security liabilities. D-Risk ID focuses on the contextual coherence of identity data (phone, email, device, address, etc.) and can cross-validate against a consortium of live, validated identities from major European retailers and banks. 

Against agentic AI 

Legitimate agents will mostly reuse known, well-behaved identities: long history, normal spend patterns, consistent device history, strong matches in consortia data. 

Malicious agents trying to mass-open accounts or test stolen identities will produce: 

• Many first-seen identities in a short window. 
• Weak or no matches to real identity graphs. 
• Synthetic patterns (odd name/email combos, phone/email geography mismatch, disposable infrastructure). 

Your scoring should treat “new identity + agent session” as high-risk by default, unless the identity is clearly rooted in your consortium / historical data. The truth is that legitimate agents actually resemble patterns and identities of legitimate users. Phew! 

Additionally, legitimate payment agents will often come from a small, stable fleet of devices with clear, contractual relationships. You can whitelist those patterns progressively once you’re confident they’re clean. 

On-site behaviour 

This is where malicious agentic AI really gives itself away. 

Tell-tale patterns for malicious agents  jump straight from entry to login/signup/payment without any browsing or hesitation; no content exploration, just form-filling. They can over-index on voucher/code entry, BNPL, gift cards, high-limit products – anything with a better payout per second. Again, they behave in very similar patterns to human fraudsters – just with greater velocity.  

By contrast, legitimate consumer agents do read product content, compare options and respect user preferences set upstream. They often return to the same merchants and categories repeatedly, with low dispute/chargeback rates over time. You can create a pattern that recognises and allows good behaviour and agents just as you do for good buyers.  

In short, your fraud stack should be able to distinguish “agent that behaves like a long-term customer proxy” versus “agent that behaves like a credential-stuffing script with a UI”. 

1. Don’t outlaw automation; classify it. 

Build explicit support for “trusted agent” traffic in your risk models instead of treating all non-human sessions as hostile. 

2. Tie everything back to a real identity. 

Lean hard on identity-graph and consortium data: if an agent is acting for identities you can’t anchor in the real world, raise the bar sharply. Talk to Oneytrust about our graph networks.  

3. Use adaptive friction, not blanket blocks. 

For “new identity + agent + high-risk product”, default to extra verification: stronger SCA, additional ID checks, or out-of-band confirmation. EU rules under PSD3/PSR are anyway pushing PSPs towards stronger screening and liability for impersonation fraud – merchants can ride that wave 

4. Align with EBA remote-onboarding guidance 

The EBA’s remote onboarding guidelines demand robust, risk-sensitive processes for online account opening, including impersonation and ID-forgery controls. That’s your mandate to deploy deeper behavioural and identity checks on agent-driven account creation without breaking legitimate customers. 

5. Instrument and rate-limit flows that agents love 

Sign-up, login, password reset, payment-instrument addition and high-risk product applications should have fine-grained rate limits and anomaly detection specifically tuned for I have tidied up the document by standardizing the titles and headers, removing all numbering, and cleaning up the content based on your instructions.

The post Agentic AI  – what you need to know in 2026  appeared first on Oneytrust.

Each return involves several parties: customer service, warehouse, carrier. Too often, these systems operate in silos, leading to delays, errors and a poor customer experience. A parcel dropped off at a collection point may remain invisible to customer service, generating calls, frustration and late refunds.

This is where Reversys comes in. Its collaborative SaaS platform puts an end to this fragmentation by orchestrating the entire return flow, from online declaration to refund. It centralises information for all parties, ensures consistency of operations and offers unprecedented agility. E-merchants can customise their policies: immediately refund a VIP customer as soon as the parcel is dropped off, or wait for verification for a high-value product. This orchestration transforms returns into a strategic lever, capable of boosting satisfaction and loyalty.

A rapidly growing phenomenon: return fraud

But behind the operational complexity lies an even greater danger: fraud. With payments becoming more secure, fraudsters have shifted their attacks to returns. This phenomenon, known as return abuse, takes various forms: returns of used or worn products under the pretext of a defect, requests for refunds for products received but declared as undelivered, or exploitation of free or extended return policies.

The figures are clear and show a worrying trend. In 2023, 13.7% of returns were deemed fraudulent, an increase of more than 10% in one year, according to Datadome. In 2024, 40% of retailers reported being victims of return abuse, according to the 2025 Global Fraud and Payments Report by Visa / Cybersource.

What is even more worrying is that the majority of companies are not prepared. In 2021, nine out of ten companies feared an increase in the risk of fraud and cybercrime, but six out of ten had not allocated a specific budget to deal with it (Fevad/Euler Hermes). In other words, the threat is known, but the means to counter it remain insufficient.

The answer: the Reversys x Oneytrust alliance

To meet this dual challenge – operational and security – Reversys and Oneytrust have joined forces. Together, they offer a unique solution that combines intelligent orchestration and predictive risk analysis.

Reversys streamlines and personalises reverse logistics, while Oneytrust, the European leader in e-commerce fraud prevention, brings its expertise in behavioural scoring and proactive detection. As soon as a return is declared, Oneytrust assesses the risk in real time. Reversys uses this information to adjust its strategy: immediate refunds for reliable customers, enhanced checks for high-risk profiles, and selection of the most secure or economical carrier depending on the context.

This approach drastically reduces fraud while guaranteeing a premium experience for honest customers. It gives e-merchants back control over their returns, transforming a process often perceived as a constraint into a lever for performance and profitability.

Did you know?

By integrating Oneytrust’s reliability with Reversys’ returns management, e-merchants can anticipate risks even before the parcel reaches the warehouse, thereby protecting their turnover and reputation.

Is your company prepared to tackle return fraud? Discover how Reversys and Oneytrust can transform your reverse logistics ! Discover the solutions offered by Reversys.

The post E-commerce returns: between operational challenges and the risk of fraud, how Reversys and Oneytrust are reinventing reverse logistics appeared first on Oneytrust.

• Easy to hijack: A stolen cookie can be enough to steal someone’s online identity.
• Rich in sensitive data: Usernames, sessions, preferences… all this information can be exploited for targeted attacks.
• Invisible to the user: Most internet users are unaware of the scope of the information stored, which facilitates fraud.

The impact on your brand

• Advertising fraud: Falsified cookies artificially inflate impressions and clicks, distorting your KPIs and increasing your costs.
• Session hijacking: Hackers can access customer accounts, generating fraudulent transactions.
• Reputational damage: A vulnerability exploited via cookies can damage your image and your GDPR compliance.

Examples of techniques used by fraudsters

• Session sniffing: Interception of cookies on unsecured websites.
• XSS injection: Malicious scripts to steal cookies.
• CSRF: Fraudulent actions performed without the user’s knowledge.
• Predictable credentials: Exploitation of weak algorithms to guess sessions.

How to protect your digital ecosystem?

• Proactive monitoring: Detecting anomalies in cookie-related behaviour.
• Advanced anti-fraud solutions: Identifying falsified or cloned cookies.
• Awareness and transparency: Informing your users and building trust.

Conclusion
Cookies are valuable tools for improving user experience and optimising marketing strategies. However, their exploitation by fraudsters poses a major risk: session theft, identity theft, advertising fraud, etc. These threats can impact customer confidence, distort your metrics and damage your image.

To remain competitive and protect your digital ecosystem, it is essential to take a proactive approach: secure your customer journeys, integrate anti-fraud solutions, and inform your users. Protecting cookies means protecting your business and your reputation.

Contact us to secure your customer journeys and build trust in your services.

The post Cookies: a marketing ally… but also a loophole for fraud appeared first on Oneytrust.

Unlike traditional identity theft, where the fraudster steals an entire real identity, a synthetic identity is a hybrid combination. It combines authentic data (in the United States, this is the social security number, while elsewhere it is usually the national identity document, date of birth and postal address) with invented information (usually contact details). This combination creates a seemingly coherent profile, especially when only the so-called authentic data is verified/analysed (as in a banking KYC process, for example).

As a result, traditional systems, which are often calibrated to validate an identity by digitising traditional practices (‘your papers, please’), miss the mark.

Why institutions are vulnerable ?
Fraudsters exploit the trust placed in partially accurate data. In the United States, with a valid official ID or minimal credit history, synthetic identities pass the initial verification stages. The fraudster can then ‘mature’ their identity: open a bank account, take out small loans, make regular payments… until they obtain a solid credit rating. This patience pays off: once credibility is established, the fraudster takes out a larger loan and then disappears. This is known as ‘bust-out fraud’.

The cost is considerable: across the Atlantic, synthetic identity fraud already accounts for several billion dollars in losses each year, according to estimates by major firms. In Europe, the figures are beginning to follow the same trend, particularly with the rise of 100% digital processes.

The limitations of traditional approaches
Traditional controls – document verification, historical scoring, ad hoc analysis of events or transactions – struggle to detect these hybrid profiles. Weak signals (email username, telephone number attributes, presence of ‘digital footprints’) can only be detected through rare expertise. Furthermore, current regulations focus almost exclusively on KYC compliance. However, control and fraud prevention are two distinct things. The former encompasses all the characteristics/steps that allow access to a service… but because all of this is exposed, it also makes it possible to understand and therefore circumvent the measures.

Towards a new approach to detection
Faced with this threat, companies must change their paradigm. The future lies in solutions capable of:

• Going beyond ‘declarative’ data by being able to add additional information.
• Cross-reference dynamic signals (browsing behaviour, machine fingerprint, location) with static identity attributes.
• Detect relational inconsistencies between different identities: shared telephone numbers, recycled addresses, accounts linked to the same device.
• Use artificial intelligence to analyse massive volumes of data/information in real time and identify patterns invisible to the human eye.

These approaches transform detection: moving from simple documentary/contextual validation to a holistic and behavioural view of the user.

A strategic challenge for digital players
Synthetic identities are not a marginal phenomenon: they directly threaten profitability, regulatory compliance and customer trust. For banks, e-merchants and insurers, investing in advanced prevention technologies is no longer an option: it is a condition for survival in the digital economy.

The post Synthetic identities: the invisible fraud that costs billions appeared first on Oneytrust.