Jumio's Privacy Notice

Inference & Profiling · Company practice

Jumio creates transaction risk calculations and scores for Customers

Section 5 lists inferences among the information Jumio may process, giving the example of a transaction risk calculation and scores. That example describes reviewing whether the IP address or other available information is known to have been used in a fraudulent transaction and providing a Customer an assessment of the likelihood that the transaction is fraudulent.

Supporting quotations

Inferences, such as a transaction risk calculation and scores (e.g., reviewing whether the IP address or other available information is known to have been used in a fraudulent transaction and provide an assessment to a Customer of the likelihood the transaction is fraudulent);

Qualifications & exceptions

• The risk-score description is an example ('such as') within the inferences category, not an exhaustive account of the inferences produced.

Not stated here

• The Notice does not state the inputs, model, or accuracy of the risk calculation, or whether the score is the sole basis for a Customer decision.

Cross-references

• qualified by (model-proposed): Jumio uses automated checks and trained-agent review, but says it does not make the final automated decision — The risk scores described as inferences inform the Customer; Jumio says it does not use automated decision-making to make the final decision with legal or similarly significant effects, and directs questions about those rights to the Customer.

Inference & Profiling · Company practice

Jumio uses automated checks and trained-agent review, but says it does not make the final automated decision

Section 14 says Jumio implements measures such as machine learning capabilities or manual review by specially trained verification agents to safeguard data quality. When processing automatically, it applies criteria such as checks on the integrity and quality of photographs and documents; extraction and analysis of text, graphical layout, documents, face photographs, biometric data and background; analysis of all steps combined, considering multiple variables, predictions and confidence values for a final score; lookups against known images and known fraudulent cases; procedures to minimize demographic bias in machine learning algorithms; and, in identity verification, comparing a selfie or video against the identification-document photograph to confirm the individual is genuinely present. Jumio says these automated checks inform the Customer, but Jumio does not use automated decision-making to make a final decision that would produce legal or similarly significant effects for the user; it directs questions about automated decision-making rights to the Customer.

Supporting quotations

Data quality and automated decision making

To help safeguard the quality of the data provided by the Services, Jumio implements measures such as machine learning capabilities or manual review by specially trained verification agents. When we process the personal information automatically, we apply the following examples of criteria:

Checks on the integrity and quality of photographs;

Checks on the integrity and recognition of documents;

Extraction and analysis of text, graphical layout, and any other available information on the documents, face photographs and biometric data, and background;

Analysis of results of all the steps combined, considering multiple variables, predictions and confidence values for a final score;

Lookups against known images as well as known fraudulent cases;

Procedures to minimize demographic bias in machine learning algorithms; and

In cases of identity verification, a selfie or video may be used to compare against the photograph on an identification document, to ensure the individual is genuinely present during the transaction.

Jumio uses these automated checks to inform our Customer, but Jumio does not use automated decision-making to make a final decision that would produce legal or similarly significant effects for you. Please contact the Customer directly with any questions regarding the rights you may have to the extent your personal information is subject to automated decision-making.

Qualifications & exceptions

• The no-final-decision statement is Jumio's own characterization; the Notice says the automated checks inform the Customer and directs the user to the Customer for questions about automated decision-making rights.
• The list of criteria is introduced by 'the following examples of criteria', so it is not exhaustive.

Not stated here

• The Notice does not describe the Customer's decision process, the role of human review by the Customer, or the accuracy of the scores.

Cross-references

• qualifies (model-proposed): Jumio creates transaction risk calculations and scores for Customers — The risk scores described as inferences inform the Customer; Jumio says it does not use automated decision-making to make the final decision with legal or similarly significant effects, and directs questions about those rights to the Customer.