A large risk model for online transactions

Elephant is a specialized AI model for online identity and payments. Trained on more than one trillion data points, it assesses transaction risk in milliseconds. Its goals are simple: faster decisions, less friction for legitimate customers, and stronger protection against fraud.

How we’re different

A model built for one problem, trained at the scale that problem demands

Payment fraud is a domain problem. It requires a model that understands not just whether a signal looks unusual, but whether it looks unusual in the context of how fraud actually behaves in payment environments. Elephant was designed around that distinction.

What that means at the model level:

Generic models apply broad logic to specific environments

Elephant is trained on payment fraud, so its scoring reflects the patterns that define legitimate and fraudulent transactions in specific payment contexts.

Signal presence alone doesn't determine fraud risk

Elephant resolves identity, behavior, and device signals together, evaluating their contextual relationship with the profile rather than their presence in isolation.

Static models fix their understanding of fraud at the point of training

Elephant is adaptive, retrained continuously to reflect each deployment's evolving fraud patterns rather than a fixed historical baseline.

Built on the data foundation a model like this requires

A large risk model is only as strong as what it was trained on. Elephant is built on Pipl's two decades of experience making fragmented global data usable in high-stakes decisions. It carries the accumulated signal depth of billions of identity connections, trillions of payment events, and twenty years of infrastructure built to make that data reliable at the point of decision.

That foundation is what allows Elephant to operate with the specificity and confidence that payment fraud decisioning demands.

Trained on over one trillion payment fraud signals

Built across connections between five billion digital identities

Twenty years of global data infrastructure underlying the model

Designed for the signal complexity of real payment environments, not training sets

Trained across the environments where payment fraud is most complex

Elephant was developed and validated across the payment environments where fraud patterns are most varied, volume is highest, and the cost of a miscalibrated model is most consequential. That breadth of training context is what gives the model its ability to perform across different deployment environments without losing specificity.

Payment Platforms

Payment platforms operate across merchant types, card schemes, and transaction behaviors where fraud patterns shift constantly. The structural challenge is maintaining scoring precision across an environment that is heterogeneous by design.

Elephant was trained on that signal complexity, giving it the context to score accurately where generic models tend to lose precision.

Trained on high-volume, multi-merchant signal environments
Calibrated to fraud pattern variance across diverse portfolios
Designed to maintain scoring accuracy as fraud behavior evolves

Model performance benchmarks across real deployment environments

In a global marketplace deployment, Elephant improved approval rates by 19.44 percentage points at the same fraud threshold, demonstrating scoring precision that generic models did not achieve in the same environment.

In a national enterprise retailer deployment, Elephant's scoring accuracy translated to an estimated $2.3M reduction in chargeback exposure.

In a global ecommerce deployment, stronger model confidence reduced manual review volume by 27% without an increase in fraud rates.

In a high-noise travel booking environment, Elephant reached 0.87 ROC-AUC, a model performance benchmark that reflects the precision of its fraud signal evaluation.