Page 36 - gs261002
P. 36
Insights and Expertise
When fraud looks While this is an extreme example, the case is a reminder
that the signals payments organizations use to
clean, assessing evaluate merchants can themselves become targets for
manipulation. Chargeback rates, business information
and transaction patterns remain valuable indicators, but
risk gets harder by themselves, they are only data points that lack greater
context.
A low chargeback rate typically provides a risk team with
useful information about a merchant. It means something
different when artificial transactions have been added
specifically to change that ratio. The same is true of a
complete merchant application. Registration documents
and a professional-looking website may satisfy individual
checks while still presenting an inaccurate picture of the
business as a whole.
That is where sophisticated risk becomes difficult. Each
piece of information may appear reasonable on its own,
while the connections between those pieces suggest
something deserves more attention.
AI can surface the problem without settling it
By Dan Frechtling AI has an important role in finding those connections
LegitScript because risk teams are evaluating signals across merchant
applications, transaction activity, websites and other
ome of the hardest merchant risk decisions sources that are difficult to piece together manually.
begin with an application that looks perfectly Automated systems can identify inconsistencies and bring
ordinary. The business is registered, the website related evidence together for review. The harder part
S is functional, ownership information appears comes when there is no familiar fraud pattern to identify.
consistent and transaction activity falls within expected
ranges. Nothing gives underwriters a clear reason to reject Sophisticated cases often involve merchants that
the application, or risk teams an obvious reason to stop have already passed several standard checks. Their
processing. The data signals don't trigger concern when documentation may be legitimate, their activity may fall
viewed within a limited scope. within normal ranges, and their websites may contain
nothing that clearly violates a policy. The concern may
Cases like these are difficult because the individual signals become visible only when several smaller discrepancies
may not look risky at all. They also raise an important are considered together.
question as payments organizations rely more heavily on
AI for fraud detection and merchant monitoring. What A business may describe itself one way during
happens when bad actors have manipulated the data a underwriting but gradually move into a different product
system is evaluating to look normal? category. Two companies presented as unrelated may
have connections that are difficult to explain. Information
Fraudsters know what risk systems look for found elsewhere online may conflict with what the
A recent federal case illustrates the problem. In August merchant told its payment provider.
2026, the owner of payment processing company CB Surety
pleaded guilty to conspiracy to commit bank fraud for None of those signals on their own necessarily means a
his role in a scheme that helped high-risk and fraudulent merchant is fraudulent. Legitimate businesses change,
merchants obtain and maintain merchant accounts (see and inconsistencies can have reasonable explanations.
https://tinyurl.com/4hu48tje).
According to the Justice Department, participants AI can help an analyst find
created sham companies, recruited straw owners, and the information that deserves
developed fake websites and contact information to make
merchants appear legitimate. The operation went further attention, but the analyst
by generating large numbers of small transactions that still has to decide what that
artificially lowered merchants' chargeback rates. More
than $111 million in transactions were processed through information means.
the scheme.
36

