AI sector rules
Counter-Fraud Framework: fraud detection systems, machine learning and artificial intelligence
SAMA Counter-Fraud FrameworkText of the Counter-Fraud Framework, October 2022, Version 1.0, published by the Saudi Central Bank (SAMA)
archived copy
Read from a public archive copy, not the publisher directly. The publisher does not serve this page to automated readers, so a direct fetch was not an option here; how we access sources. Archived . Publisher's page: https://www.sama.gov.sa/en-US/RulesInstructions/CyberSecurity/Counter_Fraud_Framework.pdfIn force. Binds private bodies.
What this law does
The Counter-Fraud Framework of the Saudi Central Bank (SAMA), October 2022, Version 1.0, defines Principles and Control Requirements for initiating, implementing, maintaining, monitoring, and improving Counter-Fraud controls within Member Organisations regulated by SAMA. The Framework defines a Member Organisation as all financial institutions or financial services providers regulated by SAMA.
The Framework is applicable to all Member Organisations operating in Saudi Arabia based on SAMA discretion, and Member Organisations required to implement and comply with it are notified by SAMA. The Framework states that it is mandated by SAMA and will be circulated to Member Organisations for implementation, and that the Member Organisations are responsible for implementing and complying with it.
The Control Requirements reflect the mandated Counter-Fraud controls that Member Organisations should consider when designing and implementing a Counter-Fraud Programme. When a Control Requirement cannot be implemented, a Member Organisation should follow an exception process that ends in a request to SAMA for a formal waiver, and approval of waiver requests is at the discretion of SAMA.
Implementation of the Framework at Member Organisations is subject to a periodic self-assessment based on a questionnaire, and SAMA reviews and audits the self-assessments to determine the level of compliance with the Framework and the Counter-Fraud maturity level of the Member Organisation. Member Organisations should at least operate at Counter-Fraud maturity level 3 or higher, which includes implementing fraud detection system capability to prevent and proactively detect fraud.
The Framework asks that a counter-fraud system that uses Machine Learning or Artificial Intelligence not be a black box and be capable of being audited, with the organisation able to test what the algorithms are designed to do and whether they are correctly implemented. The Framework defines a Black Box System as a complex system where the internal rules and mechanisms are not visible to or understood by the system owner.
The Framework defines Artificial Intelligence as the use of computer systems to perform tasks typically requiring human knowledge and logical capabilities, often in problem solving scenarios. Where the inherent risk of fraud is assessed as higher, the fraud detection standards should require additional detection controls, for example real time monitoring, additional data sources or Machine Learning models, or more stringent detection threshold criteria.
Fraud detection systems should operate 24/7 with appropriate resources in place to manage outputs on a timely basis.
Where a higher risk of fraud is identified in the Fraud Risk Assessment or higher incidences of fraud occur, Member Organisations should additionally implement system capability that includes overlaying Artificial Intelligence and Machine Learning algorithms, for example decision trees, random forests and neural networks, to enhance system decision making capability, predict the likelihood of fraud and learn from historical patterns of fraudulent and legitimate behavior.
Where a deviation from the baselined user behavior patterns is identified, a Member Organisation should either require further authentication of the user or the user's instructions, or generate an alert for further investigation to determine whether fraud has occurred.
Member Organisations should calibrate and test detection scenarios to validate that they work as designed, for example through rule logic review, threshold testing and precision and recall testing, and should implement feedback loops that review false positives, false negatives and alerts which identified fraud. Article 4 of the Saudi Central Bank Law lists, among the tasks the Bank performs, issuing the regulations and instructions relating to financial institutions and their business.
The Banking Control Law defines a bank as any natural or legal person who primarily engages in any banking business in the Kingdom.
Article 22 of the Banking Control Law provides that SAMA may, if it is established that a bank has violated any of the Law's provisions or the rules and resolutions issued for its implementation, take measures that include suspending or dismissing any of the bank's board members or staff and prohibiting the bank from extending loans or receiving deposits, on the approval of the Minister of Finance and National Economy.
Article 23 of the Banking Control Law punishes a person who violates any of the other provisions of the Law or the rules and resolutions issued for its implementation by a fine not exceeding 5,000 riyals.
What it requires