APIX LOGO
Login
help

Helpdesk

APIX LOGO
mobile menu

Global Veritas Challenge 2021

Codifying Responsible AI

DAY 00
HR 00
MIN 00
SEC 00

#veritasconsortium

Global Veritas Challenge Journey
  • Start Registration

  • Submit Proposals

  • Shortlist Finalists

  • Submit Solutions

  • Announce Winners

All timings are as per (GMT +8:00) Singapore, Beijing, Perth, Hong Kong
Organized By
In Partnership With
Powered By

About Us

About the Veritas Initiative
As part of the Veritas initiative, the Monetary Authority of Singapore (MAS), ASEAN Financial Innovation Network (AFIN) and Accenture are organising the inaugural Global Veritas Challenge 2021 – Codifying Responsible AI.

The Veritas initiative was launched in 2019 to enable financial institutions to assess their Artificial Intelligence and Data Analytics (AIDA)-driven solutions against the principles of Fairness, Ethics, Accountability and Transparency (FEAT). The FEAT principles were co-created by MAS and the financial industry to provide guidance to firms offering financial products and services on the responsible use of AIDA and to strengthen internal governance around data management and use.

[Click here] to read the whitepaper that explains the Veritas Fairness methodology.
[Click here] to watch an introduction video about the Veritas Fairness methodology
[Click here] to read the whitepaper that illustrates the methodology in two banking uses cases (Credit Risk Scoring & Customer Marketing).
[Click here] to access the two open-source codes to illustrate the Veritas Fairness methodology in two banking use cases (Credit Risk Scoring & Customer Marketing).

The concept: Fairness
Using the Global Veritas Challenge 2021 as the first step to implement Responsible AI, we will focus on the Fairness principle during this challenge.

Who can join?
We are inviting AI firms, FinTech firms, start-ups, financial institutions and tech solution providers across the globe to develop innovative technology-based solutions that can help financial service providers adhere to the Fairness principle. This would guard against bias and improve access to financial services.

How the challenge works
The Fairness problem statements have been grouped into four themes, and participating teams can select any one of the problem statements below to develop scalable and practical solutions addressing them.

Based on the proposals submitted (during registration), up to 10 teams will be selected to receive mentorship from subject matter experts and be given access to the APIX AI Sandbox (Veritas) to prototype their solutions. The APIX AI Sandbox (Veritas) is a comprehensive testing and development platform that will include data sets and APIs that teams can use as part of their solutions. From these 10 teams, three winners will be awarded S$50,000 each and the opportunity to present their solutions during the Singapore FinTech Festival 2021.

Bringing your solutions to life
On top of the cash prize, the 10 finalist teams will stand a chance to work with banks to implement their solutions. Banks may be eligible for up to S$1.5M in funding support under the AIDA grant for qualifying projects. Finalists will also have the opportunity to be part of the Veritas project in 2022 to further expand their solution with financial support.

For more information:
https://www.mas.gov.sg/news/media-releases/2021/veritas-initiative-addresses-implementation-challenges

Problem Statements

The Brief
Since 2017, responsible usage of AI has been growing extensively and 2021 will see its full expansion into the operationalisation of ethical AI principles, frameworks, and policies. To keep up, financial institutions are eagerly looking to partner with innovative companies to embed responsible AI into their business processes.

The Global Veritas Challenge 2021 is designed to embed the Veritas Fairness assessment methodology into the application of AIDA-driven solutions around four key themes. These themes were co-created with several leading financial institutions across Southeast Asia through a series of workshops to gather their inputs around industry trends, business priorities, and key challenges related to implementing FEAT principles. Findings were synthesised into the following four key themes:

  1. Product Marketing: With 43% of financial institutions and FinTech firms using AI-Enabled Marketing to create personalised customer acquisition experiences*, how do we ensure that fairness is incorporated into these AIDA solutions to promote financial inclusiveness?
  2. Risk, Compliance & Fraud Monitoring: In a global survey conducted on risk managers, 58% identify AI as the biggest potential cause of unintended consequences over the next two years**. With a growing number of financial institutions seeking to leverage AIDA solutions across multi-facets of the business including Risks and Compliance, larger volumes of data across networks, entities, jurisdiction, and industry sectors will be required to detect anomalies. How can we ensure that Fairness is incorporated into the design of these creative fraud detection AIDA models?
  3. Loan Origination & KYC: According to a survey, 52% of financial institutions are using AI to service customers in their Lending Business*. With an increasing focus on financial inclusivity, more financial institutions are leveraging alternate data sources and KYC methods to offer services to the unbanked. How do we ensure that fairness is built into such processes while providing access to financial services for the unbanked?
  4. Credit Scoring / Profiling: Financial institutions strive to automate credit decisioning, with 60% confirming their use of new/alternate forms of data for decision-making*, and processing using sophisticated AI models as well as external data for situations where historical data is lacking. This may lead to a disadvantage for groups such as small & medium-sized enterprises (SMEs). How do we minimise disparate treatments and instil confidence in machine-generated decisions?

*World Economic Forum and Cambridge Centre for Alternative Finance: “Transforming Paradigms: A Global AI in Financial Services Survey” (2020).

**Accenture: Global Risk Management Study (2019).

The submitted proposal should include the Fairness objectives and describe how the solution works with the Veritas methodology.

Based on the submitted proposals, 10 finalists will be shortlisted for the 8-week prototyping phase. The prototypes developed should demonstrate the proposed solution's application to the problem statement and showcase how it would enable financial institutions to achieve the stated fairness objectives. A set of curated APIs and data sets will be provided to the 10 finalists via the APIX AI Sandbox (Veritas). Finalists may also choose to supplement these with their own data sets.

Icon
Theme #1: Product Marketing - Problem Statement #1

Financial institutions are increasingly adopting AI-based marketing, which involves use of external customer datasets such as social media and marketing data.

How can validations and controls be built to diagnose and assess the fairness of an AIDA marketing solution which uses acquired external data?

Supporting Assets
Datasets

  1. Marketing Analytics
  2. Marketing Targets

APIs

  1. Meniga Digital Banking: Uses data to accelerate innovation and meaningful engagement by providing APIs to improve and personalise the online banking experience.
  2. Intellect Design (1): This API provides insights across various cross sell opportunities and compliance deviations.
  3. Intellect Design (2): This API module includes several endpoints enabling channels and third parties to extract various aspects of customer data.
  4. Instafin API: Lookup Client: Fetch financial institution end-client data.
  5. Intellect Design (Prospect Management): The Prospect API module includes several endpoints to allow channels, CRM systems and third parties to manage various aspects of Prospect.
  6. APIX Digital TWINN (Customer Management 2.0): Customer Management APIs allow onboarding of customers, while retrieving customer information/details from the Core Bank, and updating customer details.
Icon
Theme #1: Product Marketing - Problem Statement #2

Targeted online marketing using AI is increasingly common, where banks identify specific target segments and provide targeted marketing experiences. However, such targeted online marketing campaigns may potentially lead to bias.

How do we ensure fairness (i) during the model development lifecycle and (ii) when evaluating AIDA marketing solutions? How do we ensure that there are no intended biases against a particular segment either from the algorithm or training data?

Supporting Assets
Datasets

  1. Marketing Analytics
  2. Marketing Targets

APIs

  1. Meniga Digital Banking: Uses data to accelerate innovation and meaningful engagement by providing APIs to improve and personalise the online banking experience.
  2. Intellect Design (1): This API provides insights across various cross sell opportunities and compliance deviations.
  3. Intellect Design (2): This API module includes several endpoints enabling channels and third parties to extract various aspects of customer data.
  4. Instafin API: Lookup Client: Fetch financial institution end-client data.
  5. Intellect Design (Prospect Management): The Prospect API module includes several endpoints to allow channels, CRM systems and third parties to manage various aspects of Prospect.
  6. APIX Digital TWINN (Customer Management 2.0): Customer Management APIs allow onboarding of customers, while retrieving customer information/details from the Core Bank, and updating customer details.
Icon
Theme #2: Risk, Compliance & Fraud Monitoring - Problem Statement #1

Fraud detection solutions typically rely on historical and/or statistical datasets such as transaction patterns, demographics, and geo-location data to identify suspicious activities or behaviours. This may lead to unknown or unintended biases (e.g. a particular demographic group being flagged out).

How can fairness be incorporated into the algorithm development, testing and data handling process to ensure that there is no unintended bias against a particular group of individuals while protecting financial institutions against potential fraud?

Supporting Assets
Datasets

  1. Credit Card Fraud Detection
  2. Synthetic Financial Datasets for Fraud Detection

APIs

  1. Meniga Digital Banking: Uses data to accelerate innovation and meaningful engagement by providing APIs to improve and personalise the online banking experience.
  2. Intellect Design (1): This API provides insights across various cross sell opportunities and compliance deviations.
  3. Intellect Design (2): This API module includes several endpoints enabling channels and third parties to extract various aspects of customer data.
  4. Instafin API: Lookup Client: Fetch financial institution end-client data.
  5. Intellect Design (Prospect Management): The Prospect API module includes several endpoints to allow channels, CRM systems and third parties to manage various aspects of Prospect.
  6. APIX Digital TWINN (Customer Management 2.0): Customer Management APIs allow onboarding of customers, while retrieving customer information/details from the Core Bank, and updating customer details.
Icon
Theme #2: Risk, Compliance & Fraud Monitoring - Problem Statement #2

Use of synthetic data to supplement a lack of historical data for fraud or risk monitoring AIDA solutions is becoming common. However this increasing reliance on synthetic data may lead to risks of unintended bias.

How can fairness be incorporated (i) in the use of synthetic data and (ii) when designing and testing such AIDA algorithms? This includes generating, handling, and testing of synthetic data to prevent false positives or disadvantages to specific groups.

Supporting Assets
Data Sets

  1. Credit Card Fraud Detection
  2. Synthetic Financial Datasets for Fraud Detection

APIs

  1. Digital Asset AML (Uppsala Pte Ltd): Enables a financial application to query the Uppsala Security Threat Reputation Database (TRDB) and validate crypto address reputation in real-time. Specifically, it can be used as a plugin for Anti-Money Laundering (AML), Anti-Coin Laundering (ACL) and Crypto Asset Fraud Detection System (FDS).
  2. Intellect Design (Account Onboarding): It accepts the information relating to the holders and creates a relationship between the holders. It also checks for AML for the holders, along with performing biometric and other checks. The system captures the holders’ signatures, and the process is customised as per regulation.
  3. Solis Onboarding: This API enables KYC screening against sanction lists, PEP lists and adverse media publications.
  4. Data Zoo API: The API can be used for identity verification and PEP Screening for AML checks.
Icon
Theme #3: Loan Origination & KYC - Problem Statement #1

Many financial institutions are using historical data augmented with external data sources (e.g. social media, payment data) to create new loan origination solutions, with automated scoring and approval processes. This may lead to unknown biases, such as putting specific demographics at a disadvantage.

How can fair lending practices be ensured when using new loan origination solutions?

Supporting Assets
Data Sets

  1. Loan Data
  2. Credit Card Lead Prediction
  3. Leads Conversion Data
  4. N26 KYC Challenge

APIs

  1. Facematch (Hyperverge Fintech): Use Face Match API to verify If the selfie taken by the customer matches with the photo present in the ID card.
  2. DataZoo IDU (Singapore): The Singapore Credit service provides verification on an individual's name, DOB, and address.
  3. Handshake API: On-demand access to comprehensive corporate datasets that cover Singapore, Malaysia, and Hong Kong.
  4. Mambu Loan Transactions: Allows users to retrieve, reverse, or post a transaction for a loan account through this API.
  5. Turnkey Lender: An automated loan origination system
Icon
Theme #3: Loan Origination & KYC - Problem Statement #2

External datasets are increasingly being used for KYC and alternate identification, particularly to improve access to basic financial services for those who currently lack access (e.g. unbanked). However, the use of alternate datasets such as social media, and biometric information may contain inherent biases.

How can fairness be incorporated during the assessment, identification, and processing of alternate datasets to prevent bias during use of KYC and alternate identification AIDA solutions?

Supporting Assets
Data Sets

  1. Loan Data
  2. Credit Card Lead Prediction
  3. Leads Conversion Data
  4. N26 KYC Challenge

APIs

  1. Facematch (Hyperverge Fintech): Uses Face Match API to verify if the
    customer selfie matches the photo on the ID card.
  2. DataZoo IDU (Singapore): The Singapore Credit service provides
    verification for an individual's name, DOB, and address.
  3. Handshake API: On-demand access to comprehensive
    corporate datasets that cover Singapore, Malaysia and Hong Kong.
  4. Mambu Loan Transactions: Allows users to retrieve, reverse, or post a
    transaction for a loan account through this API.
  5. Turnkey Lender: An automated loan origination system
Icon
Theme #4: Credit Scoring / Profiling - Problem Statement #1

AIDA solutions for credit scoring and profiling require historical and reference data to accurately calibrate the credit algorithm. However, certain businesses (e.g. new businesses) which lack historical references may be disadvantaged.

How do we ensure fairness throughout the modelling development lifecycle so that certain businesses (e.g. new businesses) are not unfairly excluded during the credit scoring and profiling process?

Supporting Documents
Data Sets

APIs

  1. Lilardia Capital Credit Model: API to Access Credit Risk Rating
    Model for SME's
  2. Handshake API: On-demand access to comprehensive
    corporate datasets that cover Singapore, Malaysia, and Hong Kong.
  3. SME Financial Statement API: The financial statement API is a single point of integration that enable users to populate/produce financial statements (income statement, Balance Sheet, Cash Flow Statements) of MSMEs (formal & informal) in 22 markets.
  4. SMECreditPro API: The API is a single point of integration that enable lenders to assess the creditworthiness of formal and informal MSMEs across 22 markets (21 in Africa and Vietnam).
  5. Experian Business Directory: The API establishes credentials for existing or potential Malaysia-based customers/partners with essential information from Experian Malaysia's databank.
  6. MACH-X API: Credit Report & Rating Generation
Icon
Theme #4: Credit Scoring / Profiling - Problem Statement #2

AIDA solutions (largely based on publicly available data such as audited financial statements) are being used to support the credit evaluation and review process in financial institutions. However, new methods are being explored to augment these solutions by leveraging alternate data sets through data proxies, e.g. social media, and market connectivity.

How can fairness be incorporated during data sourcing and processing to ensure fair assessment results when using such AIDA solutions and data proxies?

Supporting Assets
Data Sets

  1. German Credit Scoring
  2. Home Default Credit Risk
  3. Credit Risk Classification

APIs

  1. Lilardia Capital Credit Model: API to Access Credit Risk Rating
    Model for SME's
  2. Handshake API: On-demand access to comprehensive
    corporate datasets that cover Singapore, Malaysia, and Hong Kong.
  3. SME Financial Statement API: The financial statement API is a single point of integration that enable users to populate/produce financial statements (income statement, Balance Sheet, Cash Flow Statements) of MSMEs (formal & informal) in 22 markets.
  4. SMECreditPro API: The API is a single point of integration that enable lenders to assess the creditworthiness of formal and informal MSMEs across 22 markets (21 in Africa and Vietnam).
  5. Experian Business Directory: The API establishes credentials for existing or potential Malaysia-based customers/partners with essential information from Experian Malaysia's databank.
  6. MACH-X API: Credit Report & Rating Generation

Prizes

What's In It For You

Up to 10 finalists and 3 winning teams will get:

A Grand Total of S$150,000 in Cash Prizes
Chance to deploy your solution in banks with funding support
Further expand your solution as part of the Veritas Project

Eligibility

  1. We welcome start-ups, small and medium-sized enterprises (SMEs) and multinational corporations (MNCs) from all geographies to participate in the Global Veritas Challenge 2021.
  2. Participants must be a legally incorporated entity.
  3. As the Hackathon will operate in English, all materials and communications submitted must be in English.

Media Centre

Introduction to the Veritas Fairness Methodology

FAQ

Am I eligible to participate in the Global Veritas Challenge?

We welcome all companies with the ability to solve any of the problem statements to apply for the Global Veritas Challenge.

Is there any geographical restriction on who can participate? 

No. The physical location of the participant does not matter. We welcome solutions from all jurisdictions and will work with the participants to facilitate an effective presentation of the solution during the judging phase. Do take note the entire challenge will be conducted virtually due to ongoing travel restrictions.

Can my company submit more than 1 proposal to different problem statements?

Yes, you may submit more than 1 proposal. However, each individual proposal should answer a different problem statement.

What is the judging process?

There will be 2 rounds of judging. The first round will select up to 10 finalists, based on the desirability, feasibility and viability of the proposed solution to the problem statement. Finalists will be selected and informed by 20 August 2021, before they proceed to prototype their proposed solution. In the second round, the top 10 finalists’ solutions will be evaluated to select up to 3 winners.

Who are the industry mentors?

The industry experts are the contributors of the problem statements or other subject matter experts who will work with you during the virtual prototyping phase to help contextualize your solution better to the industry and/or their organizations.

Is it mandatory to leverage the sandbox for our solution prototype submission?

Participating teams can leverage these technologies in the design and development of their solution prototypes. However, this is optional and not mandatory. Participants are also free to augment the provided data sets with their own externally sourced data. Participants are encouraged to propose solutions that best address the problem statements.

Will we be provided APIs and datasets to build/test our solutions?

Yes. The top 10 finalist teams will be provided with a curated list of APIs and datasets during the solutioning phase, made available on the APIX AI Sandbox (Veritas). Finalists may also choose to supplement these with their own data sets and will be given 8-weeks to build/test their solutions before Demo Day.

Will participants be able to retain all Intellectual Property (IP) rights of the proposal?

Any Intellectual Property (IP) rights of a solution remain vested in the participant that has submitted the solution. However, the participant should make available a sufficient amount of information that is necessary to fairly and accurately evaluate the solution against the evaluation criteria. The participant should also allow some details of the solution to be featured in any report that may be issued as part of the hackathon.

Contact

For Global Veritas Challenge 2021 related queries, please reach out to

veritas_challenge@accenture.com

For technical queries during registration or proposal submission process, please reach out to

apixsupport@synfindo.com