SoftTech AI Registry
Adyen · Adyen Protect

Machine Learning Fraud Risk Blocking

Model-influenced automatic payment blocking: Adyen Protect's machine learning fraud risk rule assigns each disputable transaction a risk level, and the risk engine blocks transactions classified above the merchant-configured threshold before authorization, unless an Allow rule applies.

Recorded characteristics

Function
Authority chain. (1) Machine learning model: the Machine learning: fraud risk rule evaluates the payment fraud risk of each transaction before authorization and assigns a risk level from Very Low to Very High, indicating the likelihood that the transaction will result in a fraudulent dispute. (2) Merchant-configured policy: the merchant selects the block threshold (default: high risk and above); Allow rules have the highest priority within a risk profile and can prevent a model-triggered block. (3) Adyen Protect risk engine: when a transaction is classified as riskier than the defined threshold, it is blocked before it is sent to authorization. The model influences the block; it does not independently set merchant policy or decide every aspect of payment acceptance.
Data access
The rule evaluates transaction payment properties to produce a risk classification. Fraud results, including fraudRiskLevel (veryLow, low, medium, high, veryHigh), can be returned in API responses and webhooks, but returning fraud results is not turned on by default and must be enabled by the merchant.
Actions
Can take actions
External actions
No
Human confirmation
Not required
Permission basis
Not established
Administrative control
Merchants view or change the risk threshold in Customer Area (Risk & disputes > Risk profiles). Changing the rule's settings requires the Merchant change risk settings or Risk admin role. Allow rules take precedence over Block rules. These roles govern who may change risk settings; they do not establish the technical identity that performs automatic blocking.
Default state
Not established
Availability
Requires Protect premium features; the Machine learning: fraud risk rule is added to Block rules when premium features are enabled. Not available in Protect basic. Runs on transactions that can be disputed, such as payments made with credit and debit cards.
Licensing
Premium features of Adyen Protect must be enabled.
External model or provider
Machine learning is established (Adyen's own Machine learning: fraud risk rule). The specific model architecture, version and provider are not publicly established.
Limitations and uncertainty
The specific model architecture, version and provider are not publicly established. The exact technical execution identity of the risk engine is not established. Merchant activation of Protect premium is required; it cannot be assumed that every Adyen merchant has this capability enabled. The merchant chooses or changes the blocking threshold; the model does not independently establish it. Allow rules can override a model-triggered block. Fraud-result exposure through API responses and webhooks requires configuration. The capability is documented for disputable transactions, including supported credit and debit card payments; it is not represented as covering every payment method. Default state: whether Protect premium is active for a given merchant is not established; the documented default block threshold once enabled is high risk and above.

Evidence