《SymphonyAI：2024未開發的潛力：人工智能驅動的驅動的亞洲金融犯罪合規轉型成熟度、應用與趨勢研究報告（英文版）（36頁）.pdf》由會員分享，可在線閱讀，更多相關《SymphonyAI：2024未開發的潛力：人工智能驅動的驅動的亞洲金融犯罪合規轉型成熟度、應用與趨勢研究報告（英文版）（36頁）.pdf（36頁珍藏版）》請在三個皮匠報告上搜索。

1、1Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends October 2024AI-enabled Financial Crime Compliance Transformation in AsiaMaturity,Applications&Trends Untapped PotentialResearch Report2Untapped Potential:AI-enabled Financial Crime Complianc

2、e Transformation in Asia Maturity,Applications&Trends Contents1.Executive Summary.32.Key Takeaways&Calls to Action.53.Introduction&Methodology.74.AI Deployment Maturity.95.Interview Box:Siddhant Sahai,Suncorp Bank.11 6.AI Deployment Strategies.147.Spotlight on GenAI:Opportunities&Risks.168.AI Adopti

3、on Challenges.189.Interview Box:Brent Estrella,Rizal Commercial Banking Corp.1910.AI Value Articulation.2211.Interview Box:Dr.Elea Wurth,Deloitte.2412.Regional Regulatory Snapshot.2713.SymphonyAIs Commitment to Responsible AI.3014.Final Thoughts&Strategic Recommendations.313Untapped Potential:AI-ena

4、bled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends In a world where financial crime continues to evolve at an unprecedented pace,financial institutions(FIs)must arm themselves with the most powerful tools to stay ahead.Enter Artificial Intelligence(AI)technology that

5、 revolutionises the way financial services detect risk,manage risk,ensure compliance and stay ahead of criminals,costs,and community expectations.But as exciting as the potential may be,the reality on the ground tells a more complex story.AI is not new to the financial industry;it has been deployed

6、for a long time in areas such as algorithmic trading,fraud risk,credit scoring,and chatbots.However,the advent of Generative AI(GenAI)which can create,enhance,summarise,and analyse unstructured data such as text,code,and imagesmarks a new era,expanding the number of potential use cases.The area of f

7、inancial crime is a case in point,particularly money laundering,as it represents an escalating threat,accounting for up to 6.7%of global GDP and impacting a range of financial services businesses.In this regard,bad actors hold a significant advantage over FIs due to their ability to adapt and evolve

8、 quickly.While complex structures and regulatory frameworks tend to encumber FIs,criminals are agile,rapidly adjusting their tactics to bypass safeguards.Criminal actors are already embracing AI to exploit vulnerabilities present in outdated and less sophisticated compliance systems and controls.The

9、 rapid advancement of technology,paired with slower institutional responses,creates the perfect environment for organised crime to thrive.This is driving FIs to pursue a more effective and efficient response and better ways to automate processes and generate efficiencies,including by leveraging Pred

10、ictive AI and GenAI,rather than continue with the longstanding and unsustainable industry strategy of throwing bodies at the problem.Despite the growing interest in AI,our research shows that Asias adoption of AI in financial Executive SummaryExecutive SummaryAsias adoption of AI in financialcrime r

11、isk management remainsin its early stages.“4Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends crime risk management remains in its early stages.While GenAI continues to generate significant excitement,most FIs struggle to turn this interest

12、into meaningful results.Rather,instead of delivering the expected improvements in efficiency and effectiveness,most FIs in the region remain focused on overcoming the complexities of traditional AI implementation.As a result,few have been able to fully leverage AIs potential to enhance risk detectio

13、n and streamline compliance processes,leaving much of its transformative power untapped.One of the biggest impediments to FIs applying AI in financial crime compliance processes,such as anti-money laundering(AML),is the challenge of integrating the technology into legacy systems.Many firms find it d

14、ifficult to justify the cost and complexity or to articulate the business case to their boards and senior management.Data quality and availability are also seen as significant issues,as are data privacy and protection.Firms are starting to prioritise the use of AI to help them improve data quality,r

15、ecognising that data is the lifeblood of a strong financial crime risk management programme,and the key to enabling other AI-related projects.Indeed,FIs often cite issues related to data quality,availability,reliability and integrity as the reasons that they are unable to more quickly progress their

16、 AI initiatives past pilot and experimental stages.Executive SummaryThe lack of a harmonised approach in the regulatory landscape is yet another challenge that appears to be holding back AI initiatives.Firms have insufficient clarity on regulatory expectations and struggle to align their AI projects

17、 with existing rulesets,making them cautious about full adoption.Yet,it is widely recognised that the use of AI will be a key factor in creating a competitive advantage for first movers in the years ahead.Still,the question is no longer around whether Predictive AI and GenAI will transform the finan

18、cial crime compliance function,but how.As FIs face ever-increasing alert volumes,resource constraints,and constantly evolving regulations,adopting these technologies is quickly becoming an operational imperative.Predictive AI and GenAI offer the ability to process vast amounts of data in real time,i

19、mproving efficiency and effectiveness by focusing resources on the highest-risk areas.This not only enhances compliance but also reduces operational costs,minimises disruption to customer processes,and ultimately provides a competitive advantage by enabling FIs to focus on growth while managing risk

20、s more effectively.By moving quickly from proof of concept to full-scale production,early adopters can establish themselves as leaders in efficiency and innovation before these technologies become the industry norm.FIs that seek to benefit from this first-mover advantage need to take a strategic app

21、roach to AI,including conducting assessments to determine the use cases that can generate the most value,adapting and updating their technologies infrastructure to take advantage of new innovations,and putting in place long term structures for governance,monitoring,evaluation and continuous improvem

22、ent.One of the biggest impediments to FIs applying AI.is the challenge of integrating the technology into legacy systems.“5Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Key Takeaways&Calls to ActionKey Takeaways&Calls to Action1.Predicti

23、ve AI and GenAI Adoption is in its Early Stages:Despite the potential of Predictive AI and GenAI in financial crime compliance,most FIs in Asia remain in the early stages of deployment.Many firms still grapple with legacy system integration,data quality issues,and regulatory uncertainty.2.Lack of Re

24、gulatory Clarity Continues to Hinder Adoption:A lack of harmonised regulations and clear guidance from authorities,including around ethics and their application in domains like AI,is making FIs cautious about fully embracing AI.Banks struggle to align AI projects with existing compliance rules,slowi

25、ng down adoption.3.Operational Efficiency Brings a Focus on Risk:Traditional methods of addressing financial crime are increasingly falling short in managing the speed and complexity of modern threats and allowing people the time to mitigate risk.Predictive AI and GenAI are increasingly seen as esse

26、ntial tools to enhance operational efficiency,AML effectiveness,and create the space for thinking and managing financial crime risk.4.AIs Role in Data Quality Improvement:AI technologies are being recognised for their ability to enhance data quality and governance.As FIs adopt AI,improving data mana

27、gement the lifeblood of any financial crime programme is seen as a key use case for the technology,and a crucial enabler for other AI-related projects.5.Board and Executive Sponsorship is Key:Senior executives and boards are seen as critical drivers of AI adoption.Strong leadership commitment and a

28、proactive approach to regulatory engagement are considered essential to launch new AI projects,and ultimately advance them from proofs-of-concept to larger-scale implementations.Key Takeaways from the Research:6Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,

29、Applications&Trends Key Takeaways&Calls to Action1.Move beyond customer-facing functions and explore AI-based solutions for enhancing financial crime compliance capabilities.2.Identify specific use cases where the positive impact from AI deployment is clear,such as false positive reduction,cost savi

30、ngs and efficiency gains.3.Start with pilot projects and scalable implementations before gradually expanding AIs role within the compliance ecosystem.4.Use recent financial crime incidents to demonstrate how AI can ensure more effective detection and prevention to help communicate clear value to boa

31、rds and senior management.5.Prioritise data governance to ensure high data quality standards are maintained,which is necessary to fully leverage AI technology.6.Collaborate with regulatory bodies to gain clarity and guidance on AI deployment while ensuring compliance with evolving standards.By actin

32、g on these key takeaways and calls to action,FIs can enhance their compliance programmes and ensure long-term resilience in a rapidly evolving financial crime landscape.The future belongs to those who proactively scale their AI capabilities,leverage technological advancements,and adapt to emerging r

33、isks.Calls to Action:7Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Introduction&MethodologyThe rapid digital transformation in financial services has not only opened new opportunities but also heightened risks,particularly in the area o

34、f financial crime.Bad actors have quickly adapted to the evolving landscape,continuously refining their tactics to exploit the digital environment including by misusing AI for illicit activities.As these threats grow more sophisticated,financial services businesses must move beyond traditional defen

35、ces and embrace advanced technologies like AI to stay ahead.Recent financial crime cases across Asia illustrate the pressing need for FIs to adopt Predictive AI and GenAI to detect and mitigate risks more swiftly and effectively.In Hong Kong,scammers used AI-driven deep fake video technology to impe

36、rsonate senior executives during video calls,defrauding a multinational company of USD 25 million.In Japan,during Interpols Operation HAECHI IV,authorities uncovered multiple cybercrime cases in which criminals utilised AI to bypass security measures in FIs.The specific financial losses were not dis

37、closed.Meanwhile in Australia,phishing scams powered by AI have become more prevalent.AI-enabled algorithms are being used to create highly convincing phishing emails designed to deceive corporate executives and have led to millions in financial losses.These examples illustrate the increasing sophis

38、tication of AI-based attacks,with criminals exploiting virtual communications,weak points in digital defences,and leveraging social engineering tactics to generate illicit proceeds from criminal activities.As these threats evolve with increasing speed,traditional compliance systems hampered by slowe

39、r response times and limited scalability are no longer sufficient to keep up with their increasingly complex and adaptive nature.GenAI in particular represents a powerful productivity frontier with the potential to add trillions of dollars in value across sectors,including approximately USD 200-300

40、billion for financial services.Its ability to contextually generate,analyse,and interpret vast amounts of unstructured data opens new avenues for improving operational efficiency,expanding financial crime risk coverage,and significantly enhancing detection and prevention efforts.Time is of the essen

41、ce,and FIs need to act swiftly to integrate technologies like Predictive AI and GenAI into their AML compliance and risk management frameworks.Introduction&MethodologyFIs need to act swiftly to integrate technologies like Predictive AI and GenAI into their AML compliance and risk management framewor

42、ks.“8Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Introduction&MethodologyAs regional and global regulators intensify their crackdown on financial crime,money laundering,terrorism financing,and sanctions evasion,it has become essential

43、for FIs to adopt these technologies to strengthen their defences,ensure regulatory compliance,and remain competitive.This research provides a comprehensive overview of the current state of AI deployment within FIs across Asia,with a focus on adoption maturity,implementation challenges,and a review o

44、f the factors that enable AI projects,such as organisational readiness,technological capabilities,and regulatory support.It also examines how banks measure effectiveness and return on investment(ROI),as well as how the value of AI initiatives is communicated to internal and external stakeholders.Thi

45、s report includes a dedicated section on GenAI,offering a view into current adoption trends,its potential use in financial crime prevention,and discussing some of the challenges FIs are experiencing in advancing these projects.The findings presented in this report are based on data collected from su

46、rveys and interviews conducted with 126 financial crime compliance,operational and technology practitioners across Asia Pacific between July and September 2024.The respondents represented a broad range of financial services businesses,including banking,wealth and asset management,capital markets,pay

47、ments and other financial sectors.Almost two-thirds of respondents came from firms with assets under USD 10 billion;while about 20%of the respondents came from large global firms with over USD 100 billion in assets.The majority of respondents were employed in risk,governance,financial crime complian

48、ce or technology roles,based in jurisdictions including Singapore,Australia,New Zealand,Malaysia,Philippines,Thailand,Vietnam,Indonesia,Hong Kong,and Japan,among others.Retail/Digital BankCorporate BankPrivate Bank,Wealth,AssetManagementCapital MarketsIntermediaryNon-Bank FI/PaymentsFintechOther0%5%

49、10%15%20%25%30%35%31%13%11%5%11%21%7%Type of Financial Institution9Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends AI Deployment MaturityAI Deployment MaturityDespite AI being a buzzword,the research found that its adoption and sophisticat

50、ion is still in early stages at most FIs.About 23%of respondents reported no use of AI within their organisations,while 34%assessed their sophistication in using the technology as“nascent”or“basic”.About 28%said they had“moderate”sophistication,while around 15%said they were at an“advanced”level of

51、sophistication.“The findings show a lower level of sophistication among FIs than was expected at this stage in the technologys adoption curve,”said Bradley Maclean,Co-Founder and Head of Research at Regulation Asia.“Still,it is significant that only very few firms 3.3%said they dont have plans to de

52、ploy the technology in the future,indicating the wide recognition of its transformative potential.”Sophistication of AI/ML Application in the Financial InstitutionAdvancedModerateBasicNascentNot currently in use,butplanned for the futureNot in use,no future plans14.8%27.9%29.5%4.9%19.7%3.3%0.0%5.0%1

53、0.0%15.0%20.0%25.0%30.0%35.0%10Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends AI Deployment MaturityIn AML-specific functions,deployment was also found to be lower than expected.Only a quarter of firms are actively applying the technology

54、 for AML,while another quarter are still in various stages of pilots.About 41%of respondents have yet to start on an AI adoption journey but have plans to do so in the future.A further 9.3%said they have no plans to implement AI in their AML functions.The research found that global FIs were more lik

55、ely to already be actively exploring the use of AI in AML-related initiatives compared to regional firms.Up to 55%of respondents from FIs with a global footprint said they were either actively applying AI or piloting the technology,compared to 28%of respondents at regionally-focused FIs.The vast maj

56、ority of regional firms that were not already applying AI in AML-related functions said they plan to do so in future.“There is still a strong preference for the traditional products that FIs already know,but we are seeing an increase in interest for AI-related products as the technology develops fur

57、ther and improves in sophistication,”said Craig Robertson,Financial Crime Subject Matter Expert-APAC,Financial Services at SymphonyAI.The slower adoption rate is particularly apparent in emerging markets like Vietnam and the Philippines,which appear to be prioritising AI for customer service improve

58、ments and other front office functions,leaving areas like compliance and financial crime prevention to take a back seat.“In Vietnam,out of the nine banks I spoke with,only one had its own dedicated AI development team,”Robertson said.“Their focus was entirely on improving customer experience rather

59、than compliance.”Brent Estrella,Group Chief Compliance Officer at Rizal Commercial Banking Corporation in the Philippines,told Regulation Asia that his bank has more than 50 models that extensively use machine learning,but they are largely focused on customer insights,generating sales leads and othe

60、r front office activities.The bank is however currently exploring the use of AI to trigger the CDD refresh process,based on specific events and other risk factors.The proposed model is to be trained on historical suspicious transaction reports.Full interview on page 19Yes,actively applyingYes,but on

61、ly in pilot projectsNo,but planning toNo,and there are no plans to0.0%5.0%10.0%15.0%20.0%25.0%30.0%35.0%40.0%45.0%24.6%24.6%41.0%9.8%Application of AI Technologies in AML-Related ActivitiesWe are seeing an increase in interest for AI-related products as the technology develops further and improves i

62、n sophistication.Craig Robertson,Financial Crime SME-APAC,Financial Services,SymphonyAI“11Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Q&A with Siddhant SahaiWhat have you observed in the industry regarding the use of AI and machine lea

63、rning to address financial crime risk?Sahai:There is a maturity curve in terms of the application and uses of AI and technology for financial crime related initiatives.At the high end,I have observed the use of AI-driven network graph analytics to identify hidden relationships,connected parties,crim

64、inal gangs,and syndicates and that revolutionises financial crime prevention and suspicious activity detection.Some large and international banks are using advanced analytics and machine learning for transaction monitoring to improve detection of suspicious transactions and fraud patterns,reduce fal

65、se positives,and enhance operational efficiency.Given the rising impact of scams and frauds on customers,some international banks are using AI for real time monitoring and detection of fraud risks to prevent incidents.From the perspective of a domestic bank offering vanilla products and services(e.g

66、.retail lending and deposits),similar technologies can be used but often by using established third-party technology providers instead of in-house systems for KYC,screening and transaction monitoring.Banks also apply advanced analytics for assigning customer risk ratings,i.e.to score customers based

67、 on certain factors as they emerge,e.g.transaction patterns,geographic risk,customer incidents,etc.I see an intention and hence an increasing use of machine learning in the industry to ingest large sets of data like customer transaction records to detect suspicious transactions,anomalies in transact

68、ion patterns and money mule activity which has impacted the financial services industry in recent years.Similarly,there are efforts to use natural language processing(NLP)to extract text from news articles,social media,web scraping,regulatory filings,corporate registers,and beneficial ownership regi

69、sters and then feed these inputs into machine learning models.GenAI has the ability to detect and inspect vast data sets to detect behavioural anomalies,flag suspicious transactions in real time,and apply adaptive learning,though not a lot of FIs have got to the point of full deployment.Siddhant Sah

70、ai discusses his observations on how FIs are adopting AI in financial crime risk management and some of the core challenges these initiatives faceSiddhant SahaiStream Lead-Financial Crime Oversight,Obligations and Controls Suncorp Bank,AustraliaQ&A with Siddhant Sahai11Untapped Potential:AI-enabled

71、Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends 12Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends What are the most fundamental challenges involved in deploying AI-based solutions?Sahai:The biggest underlying

72、 challenge lies in data quality,reliability and integrity,and bias in data.Good quality data is necessary for the effective use of AI and machine learning based models.Since traditional(and disparate)systems continue to be used for collecting,managing and storing customer data,FIs need mechanisms in

73、 place to ensure that the data they collect is reliable,timely,accurate and that it can be integrated across different systems.This includes ensuring data lineage,diverse and representative datasets,proper data cleansing,and extraction(without data leakages)to a data lake where it can then be used f

74、or performing advanced analytics.You also need your systems to be interoperable to ensure the smooth flow of data in the right format and while maintaining integrity.This is a significant area of work that FIs need to address in order to use analytics meaningfully to identify and detect potentially

75、suspicious transactions that warrant further investigation,develop valuable metrics and Key Risk Indicators(KRIs)for Management Information(MI)reporting,and plan assurance and testing activities.I also see a need to uplift skills and reduce the talent gap in the industry.More people are needed who h

76、ave practical experience with data,database structures,analytics and AI systems.Unfortunately,these skills and expertise are often still in short supply in the industry.Another challenge is around ensuring data protection and privacy,where there is a need to share data both internally and externally

77、,especially when offshored or outsourced.We have to ensure sensitive data is filtered appropriately,especially personally identifiable information.Data also has to be encrypted in this process and only shared on a need-to-know basis.There is an ongoing and evolving risk of cyber threats and hence en

78、suring cybersecurity is necessary,i.e.robust access controls,protecting critical infrastructure,and appropriately managing and responding to threats.Its important to ensure our data is secure and that our systems are resilient against cyber threats.What is the role of regulators in advancing AI inno

79、vations in the industry?Sahai:Currently,since there are no standardised frameworks to address ethics in AI,we need to be cautious about using both real and synthetic data for training AI models to ensure this doesnt result in model bias or discrimination.I think regulators will collaborate at the in

80、ternational level to develop such a framework and there will be more public-private alliances in this regard.Regarding regulatory guidance,some regulators like the Hong Kong Monetary Authority(HKMA)and the Monetary Authority of Singapore(MAS)have introduced AI-specific guidance to help FIs understan

81、d what they should consider from a regulatory perspective when developing use cases.Q&A with Siddhant SahaiI see an intention and hence an increasing use of machine learning in the industry to ingest large sets of data like customer transaction records to detect suspicious transactions,anomalies in

82、transaction patterns and money mule activity Siddhant Sahai,Suncorp Bank“12Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends 13Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends These

83、 currently are more principle-based,but I think this will evolve over time and clarify future approaches to address issues relating to data quality and reliability,ethics,model bias,fairness,explainability,transparency and consumer protection.I think in some jurisdictions,more could be done by regul

84、ators to give FIs flexibility to test and experiment with new technologies,such as by offering innovation sandboxes and facilitating knowledge sharing within the industry where FIs could share successful use cases.In the long run,FIs should be allowed to pool data(that meets privacy and protection s

85、tandards)for use in training AI models.This federated learning approach will genuinely help in preventing and detecting financial crime across the industry.How do you demonstrate the value of investing in AI to secure board and management support and the necessary resources to develop these initiati

86、ves?Sahai:To a degree,there is a certain level of peer pressure that can spring bank boards and senior management into action to invest in and develop AI initiatives.Leaders and senior management need to be able to demonstrate that by applying technology,they can better manage risk and regulatory ob

87、ligations effectively.Some of the other factors the board and senior management would consider are the potential long-run cost and efficiency savings,such as through a reduction in the number of tasks that have to be undertaken manually through Intelligent Process Automation.There is also value in t

88、erms of competitive advantage when one is demonstrating that you are keeping up with your peers and the industry in the application and use of AI.Where AI is entrenched,some firms may track metrics such as increases or decreases in customer complaints,a reduction in false positives,manual investigat

89、ion hours saved,or the number of fraud and scam incidents that occurred and were prevented through certain AI initiatives.Where do you see the industrys use of AI technologies is heading in the next coming years?Sahai:I think we will see move to use of conversational AI,which utilise NLP chatbots to

90、 intelligently analyse unstructured data,which can serve as educational aids for KYC and transaction monitoring analysts and investigations teams.Additionally,AI will play a bigger role in streamlining processes by using Intelligent Process Automation(e.g.document ingesting,decisioning,regulatory re

91、porting),and enabling compliance team members to concentrate on higher-level tasks.I also think federated learning will eventually become more accepted,where FIs come together to,subject to certain data protection measures being in place,jointly train AI models using their combined data sets.The who

92、le industry will benefit and learn from this,and it will help to prevent financial crime more effectively.Over time,I think the use of AI in predictive analytics will also increase,as this will enable FIs to identify potential risks ahead of time and take preventive measures.In the area of frauds an

93、d scams,there is a strong desire within the industry and among regulators to focus on real time detection and prevention rather than recovery of losses after an incident has already occurred.Disclaimer:The views expressed are personal and do not reflect the views of Suncorp Bank or any otherorganisa

94、tion.Q&A with Siddhant Sahai13Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends 14Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends The majority of respondents(78%)that said their fi

95、rms are piloting or implementing AI technologies in financial crime functions are prioritising deployment in transaction monitoring.“Transaction monitoring is a natural fit for AI due to its ability to process vast amounts of data and detect suspicious patterns more efficiently than traditional meth

96、ods,”Robertson commented,responding to the findings.The other top priority areas for AI deployment cited by respondents were KYC processes and digital verification(52%);data integrity and refresh(48%);PEPs,sanctions and watchlist screening(44%);case management and investigations(44%);transaction loo

97、kbacks(39%);and trade-based money laundering(TBML,39%).According to Maclean,PEPs and sanctions screening is seen as“low-hanging fruit”for deployment of AI technologies,while case management and investigations have always been considered one of the main“high cost areas of financial crime compliance”d

98、ue to the manual efforts needed for processes like document reviews.“AI can significantly lessen this burden and allow specialists to prioritise better and focus their time on more consequential cases,”he said.“It is not surprising to see TBML as a priority area,given the heavy burden of paper docum

99、ents imposed on trade finance practitioners.”As for performing KYC and enhancing data quality integrity,the benefits of AI in improving these processes have been widely recognised,including by regulators such as the Hong Kong Monetary Authority(HKMA).In April 2024,the regulator published a report sp

100、ecifically pointed to the ability of AI to conduct data quality checks,identify inconsistencies and errors in data,and apply automated cleansing techniques to correct these issues,remove duplicates and standardise data formats.While the report was focused on improving the performance of transaction

101、monitoring systems in the Hong Kong banking industry,it made clear that data quality enhancements was an AI use case the HKMA supported.More recently in September 2024,the HKMA issued a circular asking banks to assess the feasibility of adopting AI in their AML monitoring systems and then formulate

102、and submit an implementation plan by the end of March 2025.These were the first instructions to banks by any regulator asking them to study and plan for AI implementation.AI Deployment StrategiesAI Deployment Strategies0.0%10.0%20.0%30.0%40.0%50.0%60.0%70.0%80.0%90.0%TransactionMonitoringKYC Process

103、esand DigitalVerificationData Integrityand RefreshPEPs,Sanctions,and Watch-listScreeningCaseManagementandInvestigationsTransactionLookbacksTrade-basedMoneyLaundering(TBML)FraudPreventionEntityResolutionCybersecurityNetwork RiskThreatAssessmentAnti-briberyand Corruption78.3%52.2%47.8%43.5%43.5%39.1%3

104、9.1%34.8%30.4%21.7%21.7%8.7%Priorities for AI Deployment in Financial Crime Functions15Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends AI Deployment StrategiesWhen it comes to AI deployment,the largest number of respondents(39%)reported th

105、at their FIs take a“consultative approach”,involving multiple stakeholders from compliance,technology,and business units.Only 6%rely on a“compliance-led”approach,indicating that AI is largely seen as a business-driven initiative rather than a regulatory expectation at this stage.About 24%of responde

106、nts said such decisions were made using a“centralised decision-making”approach,involving a central decision-making body such as a dedicated AI governance board or executive team.Just 6%reported using“decentralised decision-making”,where individual departments or teams can make AI deployment decision

107、s independently and within agreed guidelines.About 24%of respondents said they use a“hybrid approach”with both centralised and decentralised aspects.The research highlighted the critical role of the board and senior management in driving AI adoption,with 40%of respondents saying their top leadership

108、 are the“primary advocates”for AI adoption,actively setting strategies and priorities.A further 35%of respondents said the board and senior management play an“active supporting”role through resources and policymaking,while 18%said they play a“consultative”role,providing advice and guidance on key de

109、cisions.Around 10%of the respondents said the top leadership has“limited”or“no involvement”in AI adoption decisions.“To a degree,there is a certain level of peer pressure that can spring bank boards and senior management into action to invest in and develop AI initiatives,”said Siddhant Sahai,Stream

110、 Lead for Financial Crime Oversight,Obligations and Controls at Suncorp Bank in Australia.“Leaders and senior management need to be able to demonstrate that by applying technology,they can better manage risk and regulatory obligations effectively.”“Some of the other factors the board and senior mana

111、gement would consider are the potential long-run cost and efficiency savings,such as through a reduction in the number of tasks that have to be undertaken manually through Intelligent Process Automation,”he added.“There is also value in terms of competitive advantage when one is demonstrating that y

112、ou are keeping up with your peers and the industry in the application and use of AI.”Full interview on page 11Another research participant said it is much more difficult to advance AI-related initiatives when board involvement is limited or non-existent,and that this can lead to“disillusionment”in m

113、any cases.“FIs that fail to align AI strategies with broader business goals risk stalling their AI projects and feeding into the cycle of unmet expectations,”he said,speaking from a Singapore bank.Decision-Making Approach for AI Deployment and Governance Structure24.2%6.1%24.2%39.4%6.1%0.0%5.0%10.0%

114、15.0%20.0%25.0%30.0%35.0%40.0%45.0%CentralisedDecision-MakingDecentralisedDecision-MakingHybrid ApproachConsultativeDecision-MakingRegulatoryCompliance Led16Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Spotlight on Generative AI:Opportu

115、nities&RisksSpotlight on GenAI:Opportunities&RisksIn the financial crime space,GenAIs ability to contextually process,summarise,and analyse vast amounts of unstructured data positions it as a valuable tool for addressing long-standing industry challenges.Despite this optimism,adoption remains in its

116、 early stages.Through this research project,early adopters of GenAI in AML processes reported mixed results.Improvements in areas like suspicious activity reporting(SARs)and data analysis were cited,but some respondents pointed to persistent challenges in scaling these technologies for real-time com

117、pliance operations.One regional bank reported faster SAR generation using GenAI,but that manual verification of critical information sometimes found errors including“hallucinated”text,causing delays in urgent cases,and lowering confidence in the technology.Yet,another respondent working for a large

118、payments firm described GenAI as being“particularly helpful”when large volumes of SARs need to be filed,as the technology can first identify all the SARs with similar characteristics,and then write standardised narratives for all of those“similar”cases before sample checks are conducted and the SARs

119、 are submitted.15%of respondents said they were actively applying GenAI in AML processes,while a further26%said they were doing so in pilot projects.17Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Spotlight on GenAI:Opportunities&RisksTh

120、e same respondent also discussed the use of GenAI to“cut through the noise”to identify negative news or adverse media articles that are“highly relevant”to case managers and investigators,and synthesise this into information that can be more quickly and easily digested.He said this is being trialled

121、at several FIs,along with a number of use cases involving the use of GenAI for document reviews as part of the CDD process.In the research,close to 15%of respondents said they were actively applying GenAI in AML processes,while a further 26%said they were doing so in pilot projects.Almost half(49%)s

122、aid they were planning to use the technology in future,while just 10%said they had no plans to do so.Globally,this trend is similar.According to the World Economic Forum,60%of FIs are piloting or deploying GenAI,yet only around 16%report widespread AI use in compliance operations.The decision to del

123、ay adopting GenAI can stem from a mix of technological,financial,regulatory,and organisational challenges.For the respondents already using GenAI,the most frequently cited use case was for transaction monitoring,followed by alert processing;KYC/CDD processes;and customer risk scoring.For some firms,

124、the decision is based on the perceived benefits compared to other projects.For instance,RCBCs Brent Estrella in the Philippines said his bank has“thought about”using GenAI for sanctions screening and writing STR narratives,but that this is“not yet an immediate priority”.This is largely due to manage

125、able alert volumes and a focus on optimising other priority areas first.Full interview on page 199.8%14.8%26.2%49.2%0.0%10.0%20.0%30.0%40.0%50.0%60.0%Yes,actively applyingYes,but only in pilot projectsNo,but planning toNo,and there are no plans toUtilisation of GenAI Technology in AML ProcessesThe d

126、ecision to delay adopting GenAI can stem from a mix of technological,financial,regulatory,and organisational challenges.“18Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends AI Adoption ChallengesAI Adoption ChallengesUnderstanding the challe

127、nges many FIs face when applying AI in AML provides insight into why adoption is not as high as expected.In the research,the biggest challenges reported by FIs were“Integrating AI with Legacy AML Systems”and“Data Quality and Availability”both cited in a multi-select question as the top issues by nea

128、rly 60%of respondents.SymphonyAIs Robertson said he has observed many firms struggling to justify investing in AI when they are still reliant on legacy infrastructure.“Theres a real risk of over-promising and under-delivering when it comes to AI,”he said.“Many firms still see AI as a long-term proje

129、ct,especially when they hit the complexity of integrating AI into legacy systems.Data integrity also continues to be a major focus in our conversations.Organisations should absolutely think long-term but they should also be aware how easily effective AI can be when implemented to deliver a low-impac

130、t outcome and without much process upheaval in the short term.”The data-related challenges cannot be understated,as AI is unable to effectively function without good-quality data.Suncorps Sahai said data quality,reliability and integrity is the“biggest underlying challenge”when it comes to AI adopti

131、on,and that this is a“significant area of work”that every FI needs to address to effectively and meaningfully use the technology.Another respondent spoke about having to“start from scratch”to create a data governance framework,because his bank did not have one.One respondent described a two-year-lon

132、g process that it took for his firm to reach a point where its data could be used in an AI project.Some respondents recognised that AI itself can help address these challenges by automating data cleansing and structuring processes,reducing the time and effort needed.AI tools can detect inconsistenci

133、es and anomalies,speeding up the data preparation process and allowing FIs to deploy AI more quickly.This dual role of AI both as a user and enabler of high-quality data demonstrates the technologys potential to overcome key barriers to adoption.Model explainability,data privacy and protection,model

134、 development and training,and regulatory issues were also among the top challenges cited by respondents.Around 43%said data privacy and protection rules were among their top challenges,while 38%cited difficulties ensuring compliance with existing regulatory requirements for FIs;and 28%said they stru

135、ggled to gain regulatory or supervisory approval for their AI initiatives.Top Challenges of Implementing AI for AML0.0%10.0%20.0%30.0%40.0%50.0%60.0%70.0%58.6%58.6%46.6%43.1%41.4%39.7%37.9%36.2%29.3%27.6%27.6%25.9%15.5%Integrating AIwith ExistingAML SystemsData Qualityand AvailabilityModelExplainabi

136、lityData Privacyand ProtectionComplexity ofAML SchemesModelDevelopmentand TrainingEnsuringRegulatoryComplianceScaling AISolutionsRecruitmentand Retentionof AI TalentData SharingChallenges(Non-Privacyrelated)GainingRegulatory andSupervisoryApprovalBias andFairnessPerception fromFirst andSecond Lines1

137、9Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Q&A with Brent EstrellaAre you currently adopting AI in the financial crime space or other areas?Estrella:We have more than 50 models that extensively use machine learning.These are largely

138、focused on customer insights,generating sales leads and things like that.So far there have been fewer AI initiatives in the fraud and financial crime space.But one of our first steps in this space is to develop a model that will allow us to operationalise a trigger event-based,dynamic CDD refresh pr

139、ocess.This is partly driven by regulatory requirements to perform ongoing CDD.But its also driven by resources in terms of how big the teams are that can actually undertake CDD reviews.I am not a believer of the current approach most banks take of performing periodic CDD reviews at set intervals,as

140、these tend to only to come up with the same conclusions that youve had at the last time that you did the review.In fact,about two years ago we estimated that we would need to hire more than 500 CDD analysts if we were to take the traditional approach of conducting periodic reviews at fixed intervals

141、.We wanted to move away from just throwing bodies at this task and come up with something which is trigger based.Two years down the line,were now at the point where our compliance and data science teams are exploring how we can build a model that will help us deliver that.How do you envision that us

142、ing AI in your ongoing CDD process will help the bank?Estrella:Part of the discussion weve had is that the customer risk rating model weve had in place for the past six years is not quite as advanced as we would like it to be.We decided we wanted to use AI in order to scale up our customer risk asse

143、ssments.For example,CDD refresh triggers could be the clients product utilisation history and risk factors that become evident during the relationships.We decided to look at ways to use partially redacted suspicious transaction reports to train an AI model to look into the demographics and attribute

144、s of Brent Estrella discusses RCBCs approach to the use of AI and machine learning and how decisions are made to develop these initiatives.Brent EstrellaGroup Chief Compliance OfficerRizal Commercial Banking CorporationPhilippinesQ&A with Brent Estrella19Untapped Potential:AI-enabled Financial Crime

145、 Compliance Transformation in Asia Maturity,Applications&Trends 20Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends customer accounts that we have investigated and found to be suspicious.We believe this will give us a better way of risk-rati

146、ng customers from an AML perspective.What initial foundational challenges did you have to address to get to a point where you were able to manage AI models?Estrella:We had to start with the most fundamental issues,like establishing a data governance framework.We have since written out a document cle

147、aring stipulating our data governance framework,covering all the different kinds of data and the roles of data owners,for example.This serves as the foundation for us to harness the power of data down the line.Previously,a significant amount of the data was unstructured,much of it on paper and not d

148、iscoverable.Getting this into a system to make it usable in our models was quite a challenge.The process of establishing a data governance framework was not easy.Just agreeing on the responsibilities across the organisation was in itself like a Herculean task.But ultimately it was necessary to go th

149、rough that painful journey to make sure that it was done right.How are decisions made regarding the deployment of AI initiatives?Estrella:We are fortunate to have a CEO who is a firm believer in the power of data and AI.And he has people like myself and my other peer group heads who likewise believe

150、 the same.So there is absolute alignment between the department heads and management.Not to be forgotten is the board.We have a very supportive board in terms of us going down this route.One of our independent directors happens to be the head of the Centre for AI Research in the Philippines,which al

151、lows for very good alignment within management.Of course,there tends to be a natural bias for business related uses of AI or machine learning rather than compliance.So this means that using such technologies in compliance functions can take longer.But Ive found that it is important for people at my

152、level to assert ourselves and communicate our interest in data and technology.One thing that will help us drive our agenda is having more specialised technical skill sets within the compliance function.This would allow us to pursue this journey from within,rather than having to compete for resources

153、 from our banks data science team.How do you articulate to the board the value of investing in AI initiatives for risk and compliance functions?Estrella:As a starting point,our board believes in the principle that compliance is a competitive advantage and is very much engaged on risk and compliance

154、matters.With that alone,you can really go quite far in terms of what investments the bank will actually make into compliance.Q&A with Brent EstrellaWe wanted to move away from just throwing bodies at this task and come up with something which is trigger based.Brent Estrella,RCBC“20Untapped Potential

155、:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends 21Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends In terms of metrics,its quite difficult to measure the benefits of AI in terms of additional reven

156、ues or higher profits,especially in compliance.But ultimately,our goal is to obtain a perfect supervisory rating,which means that our board is always prepared to make the necessary investments to get us there.Have you started exploring the use of GenAI within your organisation?Estrella:We have thoug

157、ht about using GenAI for sanctions screening and for writing STR narratives,as this would free up staff and speed up case management system and decisioning.But we feel this is not yet an immediate priority right now,because our alert volumes are still at a level that our staff can manage well.If you

158、 talk to a global bank with significantly greater volumes of alerts and regulatory pressures from different corners of their operations,then its probably going to be a very different conversation.What drives decisions and appetite to spend on GenAI adoption?Estrella:Its really a revenue conversation

159、.For now,GenAI is really more of an optimisation,and there are many other areas for optimisation that are seen as higher priorities.But to be fair,if there was a use case for GenAI that we really felt we needed,for example to satisfy a regulatory concern,then our bank would not hesitate to make that

160、 move.What steps is Bangko Sentral ng Pilipinas(BSP)taking to ensure AI is adopted responsibly in the banking industry?Estrella:The BSP conducted a thematic review to understand how some of the banks are using AI.Our bank was included as one of the participants in this review.Based on the findings f

161、rom this review,the regulator is developing a framework for the industry focusing on the use of AI.This is currently expected to be released before the end of this year.Beyond that,whenever we get a sense that there could be a regulatory issue,we are able to adopt a sandbox approach with our regulat

162、ors.This means going to talk to the regulator and explaining what we are planning,and asking them to allow us to experiment within a sandbox.The process of establishing a data governance framework was not easy.Just agreeing on the responsibilities across the organisation was in itself like a Hercule

163、an task.But ultimately it was necessary to go through that painful journey to make sure that it was done right.Brent Estrella,RCBC“Q&A with Brent Estrella21Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends 22Untapped Potential:AI-enabled Fin

164、ancial Crime Compliance Transformation in Asia Maturity,Applications&Trends Understanding the cost and benefit to the business is an important step for FIs in any AI journey,particularly in an economic climate that has firms becoming increasingly cost-conscious.In many cases,where FIs are deploying

165、AI,it is in revenue-generating areas such as sales and marketing rather than compliance.Therefore,a strong business case has to be communicated to secure board and senior management buy-in and the resources necessary to pursue AI initiatives for AML-related functions.In the research,respondents were

166、 asked about the metrics they use to measure the effectiveness of AI applications in AML operations and how they communicate value of these initiatives to their top leadership.The most commonly(61%)cited metric was the false positive rate.High false positives generated by transaction monitoring and

167、screening systems have long been a burden on FIs,particularly those with large global footprints and significant cross-border business.For most FIs,false positives take significant time and manual effort from case managers and investigators to clear.Further,false positives can create negative experi

168、ences for customers,slow business growth,and hinder efforts to address genuine money laundering concerns.FIs that can make the case for lowering false positive rates using AI are therefore more likely to be successful in securing board and management buy-in.AI Value ArticulationAI Value Articulation

169、The most common metric to communicate value of AI applications in AML operations was the false positive rate,cited by 61%of respondents.23Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends AI Value ArticulationAccuracy rate(52%)was the second

170、 most commonly cited metric used to demonstrate the effectiveness of AI projects,followed by efficiency improvements(46%),detection rate(44%),and cost savings(43%).These metrics,when shown to be improving,all mean that compliance officers,case managers and investigators are freed up to focus on high

171、er-value activities and cases.While AML compliance functions have to demonstrate cost-savings,front office functions demonstrate the revenue opportunity,arguing a more powerful driver of board and senior management support.According to Regulation Asias Brad Maclean,those working in AML roles should

172、emphasise the importance of financial crime compliance when communicating the value of their AI projects,which can include discussions on the legal,regulatory and reputational risks associated with non-compliance and detection failures.Many respondents in developed markets like Australia,Singapore,a

173、nd Hong Kong,where the cost of compliance is considered significant compared to less developed markets,report that they seldom experience pushback from boards and senior management for AI projects aimed at enhancing their financial crime risk management capabilities.In part,this is due to support fo

174、rm regulators in terms of regulatory guidance and the high priority placed on guarding against money laundering and other types of illicit finance activities.“In markets like Hong Kong,compliance costs are sky-high,and AI is seen as a way to reduce those costs,”said Robertson.“In some other markets,

175、weve seen firms struggle to align their AI objectives due to a lack of regulatory guidance.The technology moves faster than the regulations,so firms can be hesitant to invest heavily without clearer guidance from regulators.But it is imperative that firms prioritise and communicate the importance of

176、 enhancing the efficiency and effectiveness of their financial crime controls and systems.”In markets like Hong Kong,compliance costs are sky-high,and AI is seen as a way to reduce those costs.Craig Robertson,Financial Crime SME-APAC,Financial Services,SymphonyAI“60.7%52.5%45.9%44.3%42.6%32.8%27.9%1

177、8.0%16.4%13.1%0.0%10.0%20.0%30.0%40.0%50.0%60.0%70.0%False PositiveRateAccuracy RateEfficiencyImprovementDetectionRateCost SavingsAudit andTraceabilityComplianceRateFalseNegative RateScalabilityIndexCustomerImpact ScoreMetrics Used to Measure Effectiveness of AI Applications in AML24Untapped Potenti

178、al:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends How mature is Australias financial services sector when it comes to implementation of AI?Wurth:Australia is generally lagging globally when it comes to AI innovation and deployment for value.Financial servic

179、es demonstrate the full range of ambition,where some see AI as integral to their future and others retain a bricks and mortar strategy.Certainly some of our larger banks are highly progressed in this space,but as an industry,I would say that financial services is not a front runner in either maturit

180、y of implementation or AI governance.What are the main use cases FIs are prioritising when it comes to financial services?Wurth:In September,the government proposed mandatory guardrails for AI in high-risk settings,which essentially gave us an indication of what AI regulation is going to look like f

181、or high risk AI.Theres certainly some FIs looking to implement AI for AML,but AML and many of the customer facing AI solutions would be considered high risk under the Australian definition,so there is a higher requirement for risk management and governance around these kinds of use cases.Weve found

182、theres a bit of reticence in the financial industry to move towards customer facing or high risk AI models without really understanding what the regulation will be,and how to effectively manage the risk of those models.So what we tend to see is that early AI deployment within financial services is i

183、nternal facing,like back office operations.These deployments tend to focus on productivity and efficiency gains internally in back office rather in customer facing functions at this point.What are the key areas FIs should focus on when it comes to AI governance?Wurth:When we talk about governance,we

184、re talking about frameworks set up to protect the organisation,their customers,society,and ensure they are adhering to the appropriate emerging regulations,corporate policies,and societal expectations,for example.Dr.Elea Wurth discusses the maturity of AI in Australias financial industry and how fir

185、ms should go about developing their governance frameworks.Dr.Elea WurthLead Partner Trustworthy AI Strategy,Risk&TransactionsDeloitte,AustraliaQ&A with Dr.Elea WurthQ&A with Dr.Elea Wurth24Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends 25

186、Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Firms should make sure theyre implementing a governance system that gives visibility and control of AI procurement,deployment,and use within the organisation,covering the users of the AI syst

187、em all the way up to the board.When youre looking at developing those governance structures,the focus should be on ensuring that your AI is ethical,is lawful,and is accurate.Accuracy refers to the AI being technically robust,that its doing what you think its doing.It is incredibly important to be ab

188、le to define your ambition and AI strategy upfront to help define what your AI governance needs actually are.Some organisations have very simple ambitions,such as to allow their staff to use publicly accessible GenAI tools like ChatGPT.In these cases,you dont need an enormous in depth,complex govern

189、ance structureHow do firms with larger AI ambitions tend to go about developing their governance frameworks?Wurth:Organisations with larger ambitions may go ahead and develop AI use cases and build out their POCs and POVs.In some cases,it is only when they actually get to the point of wanting to dep

190、loy into production that they suddenly realise they dont have any way to manage the risk.This can be a blocker to achieving value from AI,because no executive or board is going to provide the approvals or the funding to go into production if they dont feel comfortable that the AI programme is being

191、effectively risk managed and appropriate safeguards are in place.Other organisations are more risk averse,so they tend to focus on developing their AI risk frameworks and governance before they start investing in innovation,so they can feel comfortable in that investment and innovation.Thats another

192、 extreme.The most effective organisations tend to develop their AI strategy and governance frameworks in unison.By the time they get to a point where theyve gone through their development and experimentation phase,theyve also already built out safeguards and risk management along the way,so theyre i

193、mmediately ready to move into productionisation and value achievement.How do firms go about measuring and demonstrating that their AI programmes are achieving value?Wurth:Its incredibly important to have value measurement in place upfront.Many organisations will go forth and develop and invest in AI

194、 programmes without really understanding the value that they need to be achieving.Q&A with Dr.Elea Wurth25Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends So what we tend to see is that early AI deployment within financial services is inter

195、nal facing,like back office operations.These deployments tend to focus on productivity and efficiency gains internally in back office rather in customer facing functions at this point.Dr.Elea Wurth,Deloitte“26Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Ap

196、plications&Trends Organisations should not be looking to procure,develop,or deploy AI just for the sake of having AI.There needs to be a real focus on what is the value metric for your AI programme.This needs to be set clearly upfront,and there should be regular check-ins to make sure those value me

197、trics are being achieved.So when were looking at AI strategy and governance,value management and setting those metrics upfront is absolutely a key component of any AI programme.What should governance look like for firms that are using AI developed by third-party firms rather than in-house?Wurth:Even

198、 if youre buying off the shelf and youre going with a major provider,as the deployer of that AI,you are responsible for any risk that is realised from the deployment.So you still need to have a governance framework,the same as you would if you were developing your AI internally.No matter where you a

199、re on that journey,you need to have your own risk assessment in place,and ensure that you manage the risks appropriately.You would still need to right-size your AI risk management governance,so there would be some differences for those developing AI internally vs deploying from external providers.Q&

200、A with Dr.Elea Wurth26Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Organisations should not be looking to procure,develop,or deploy AI just for the sake of having AI.There needs to be a real focus on what is the value metric for your AI

201、 programme.This needs to be set clearly upfront,and there should be regular check-ins to make sure those value metrics are being achieved.Dr.Elea Wurth,Deloitte“27Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Regional Regulatory Snapshot

202、Regional Regulatory Snapshot1.Singapore:Leading in AI Governance with a Balanced ApproachSingapore has positioned itself as a frontrunner in AI regulation,led by the Monetary Authority of Singapore(MAS),which emphasises ethical AI adoption through its Fairness,Ethics,Accountability,and Transparency(

203、FEAT)principles.The Veritas Initiative further helps FIs ensure that their AI processes comply with governance and ethical standards.Singapores AI Governance Framework,driven by IMDA in coordination with industry via the AI Verify Foundation,also promotes transparency and security,setting the benchm

204、ark for balancing innovation with regulatory oversight.2.Hong Kong:Emphasising Transparency and AccountabilityHong Kongs regulatory landscape focuses on ensuring transparency and accountability in AI adoption,particularly through frameworks like the Personal Data(Privacy)Ordinance(PDPO)and the initi

205、atives led by the HKMA.The AI Supervisory Sandbox allows FIs to test AI-driven anti-fraud and AML technologies in a controlled environment.The HKMA has recently become the first regulator to instruct FIs to assess the feasibility of adopting AI in their AML monitoring systems and then submit impleme

206、ntation plans to the regulator.The regulatory landscape for AI in Asia is rapidly evolving,with various countries taking significant steps to define how AI should be deployed,particularly in financial services.While the enthusiasm for AIs transformative potential is evident,the lack of unified frame

207、works across the region presents both challenges and opportunities.28Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends 3.Japan:Cautious but Focused on AccountabilityJapans approach to AI adoption is more conservative.The Financial Services A

208、gency(FSA)emphasises transparency and accountability,ensuring AI does not compromise consumer protection or contribute to discrimination.Although AI use in financial crime compliance is still in its early stages,Japans ethical guidelines and clear privacy laws ensure a secure regulatory foundation a

209、s FIs gradually adopt the technology.4.Australia:Mandatory Guardrails for AIAustralia is proposing to introduce mandatory AI guardrails and voluntary standards to ensure ethical AI deployments in high-risk sectors,including financial services.The Australian Governments AI Ethics Framework emphasises

210、 consumer protection and system integrity,with guidelines focused on transparency and risk management.Released in 2024,Australias regulations are a significant step toward integrating AI with clear governance structures.5.Malaysia:Plans for AI Hub with a Focus on RegulationMalaysia has announced pla

211、ns to rapidly position itself as a key player in AI development,with its recent initiative to establish itself as a regional AI hub.The strategy focuses on driving AI adoption across industries,particularly financial services.Regulatory frameworks are being shaped to align with global standards and

212、support AI innovation while ensuring risk management around data privacy,transparency,and accountability.6.Vietnam:Draft Law on Digital Technology and DataVietnam is advancing its AI and data governance regulatory framework by introducing laws on digital technology and data,focusing on enhancing dat

213、a privacy,security,and transparency.A notable feature of these laws is data localisation,requiring certain types of data collected in Vietnam to be stored locally,which may affect cross-border AI applications.In June 2024,Vietnams Ministry of Science and Technology issued nine principles for researc

214、h and development of responsible AI systems and guidelines for implementation.7.China:First Mover with GenAI-Specific FrameworkChina has positioned itself as one of the first markets to move on GenAI regulation,introducing GenAI-specific frameworks in 2024.These Regional Regulatory Snapshot29Untappe

215、d Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends frameworks align with the countrys broader data security,data privacy and cybersecurity laws,focusing on content control and accountability in AI deployment.Chinas approach includes strict guideline

216、s for AI-generated content and mandates clear human oversight in high-risk sectors,making it one of the first countries to comprehensively regulate GenAI.8.Indonesia:Balancing Innovation with Strong Regulatory OversightIndonesia is actively developing its AI regulatory framework to balance innovatio

217、n with strong oversight.The Ministry of Communication and Informatics(Kominfo)has introduced guidelines emphasising ethical AI adoption,transparency,and accountability.Indonesias National AI Strategy focuses on driving AI innovation while ensuring robust governance structures are in place.The countr

218、y is also exploring the establishment of an AI sandbox to allow FIs to test the technology in a controlled environment,and preparing to issue a ministerial regulation on AI.9.Thailand:Emphasising Data Privacy and SecurityThailand is enhancing its AI regulatory landscape with a strong focus on data p

219、rivacy and security,with the Electronic Transactions Development Agency(ETDA)and the National Digital Economy and Society Committee(NDESC)leading the effort.Thailands Personal Data Protection Act(PDPA)ensures strict data protection measures,while the AI Governance Framework emphasises transparency,f

220、airness,and ethical AI deployment.Thailands government is also encouraging the development of AI ethics committees within organisations to oversee AI implementation and compliance,with policymakers planning to draw up new AI rules to prevent the ethics violations.10.Philippines:Building AI Capabilit

221、ies with a Focus on Emerging RegulationThe Philippines is in the early stages of AI development,with several pilot projects in the financial services sector focusing on fraud detection and compliance.Broader integration across other industries is still in its infancy,with significant investment stil

222、l needed required to support large-scale AI adoption.The Philippines is actively working on draft legislation focused on areas such as data privacy,cybersecurity,and the ethical use of AI.The central bank is meanwhile working on AI guidelines for the banking sector.The Philippines has also proposed

223、the introduction of a comprehensive AI regulatory framework for the ASEAN region by 2026,covering cybersecurity and GenAI,among other areas.Regional Regulatory Snapshot30Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends SymphonyAIs Five Resp

224、onsible AI Principles are:SymphonyAIs Commitment toResponsible AISymphonyAIs Commitment to Responsible AISymphonyAI is a leading enterprise AI SaaS company for digital transformation across the most critical and resilient growth verticals including financial services.The firm is focused on ensuring

225、responsible AI practices in its enterprise AI solutions,placing emphasis on maintaining data integrity,mitigating bias,and promoting fairness in AI systems.This approach addresses key regulatory considerations and enables real-time feedback,which is critical in regulated industries.By integrating hu

226、man oversight,SymphonyAI aims to ensure AI decisions are reliable and aligned with operational standards.By adhering to these principles,SymphonyAI ensures that its AI technologies are not only innovative and effective but also secure,and adaptable to evolving regulatory frameworks,standards,and bes

227、t practices.This commitment fosters trust and confidence among stakeholders,supporting responsible and sustainable AI adoption across industries.1.Accountability:AI systems should be designed and operated in a way that ensures stakeholders can hold product users responsible for their actions and dec

228、isions.2.Transparency:Their decision-making processes and potential biases must be clear,understandable,and open to customers,developers,and regulators.3.Reliability&Safety:They should behave predictably and within customer and industry acceptable risk thresholds to ensure customer trust and confide

229、nce.4.Security:Security of information and systems,of both SymphonyAI and its customers,must be maintained and sustained throughout the full AI system and data lifecycles.5.Privacy:SymphonyAI adopts the principle of Privacy by Design.All AI systems designed,developed,implemented,maintained and decom

230、missioned must meet the requirements of applicable laws,regulations and industry standards.31Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Final Thoughts&Strategic RecommendationsFinal Thoughts&Strategic RecommendationsGenAI adoption in

231、financial crime compliance across Asia Pacific FIs has been slow,with most firms still navigating the complexities of traditional AI implementations.Legacy AML systems,data quality issues,and a fragmented regulatory landscape have further hindered progress.While there is growing interest in AI,the m

232、ajority of FIs remain in the early stages of deployment,focusing primarily on customer-facing applications rather than compliance.However,the urgency to embrace AI in compliance functions has never been greater.Recent high-profile financial crime cases have exposed vulnerabilities in traditional com

233、pliance systems,highlighting opportunities where AI and GenAI could have played a more significant role in identifying and mitigating risks earlier.For financial crime compliance teams,leveraging these cases to outline the clear gaps and inefficiencies in current systems can drive stronger buy-in fo

234、r AI programmes.In Japan,a criminal organisation called the Rivaton Group was caught using 500 shell companies and 4,000 bank accounts to launder JPY 70 billion(USD 470 million)of illicit proceeds in the first half of 2023.The banks that opened those accounts were not able to detect links between th

235、e accounts or identify that transfers made to online casinos and overseas corporate accounts were suspicious.The FSA issued directions to FIs in September 2024 outlining a series of measures they are expected to take to improve customer verification,behaviour analysis,transaction monitoring,and dete

236、ction scenarios,among other areas.In Singapore,ten foreigners were arrested in August 2023 and charged for money laundering related crimes.About SGD 3 billion(USD 2.3 billion)in assets Final Thoughts:32Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applicati

237、ons&Trends are involved in the case.While the ten individuals were ultimately convicted and jailed,multiple Singapore banks have been under scrutiny for failing to detect false documents during onboarding.The MAS has directed banks to improve their processes for verifying sources of wealth(SOW),and

238、plans to impose fines and other punitive measures on some of the banks that had dealings with the ten convicts.These two high-profile financial crime cases identify system and control deficiencies that can already be enhanced with current AI technologies,including customer verification,behaviour ana

239、lysis,transaction monitoring,and false document detection.With regulators already directing FIs to fix these areas,the case for leveraging AI to gain efficiency in the process makes for a stronger case when trying to secure buy-in from stakeholders.By demonstrating how AI can address critical gaps s

240、uch as enhancing detection accuracy,reducing false positives,and speeding up response times FIs can turn these challenges into opportunities for technological advancement and a more refined approach to risk management.This improves operational efficiency and deepens the ability to understand,anticip

241、ate,and mitigate risks,offering a more proactive stance in the ever-evolving landscape of financial crime.The ability to leverage AI for a better grasp on risk factors transforms compliance efforts into a strategic asset for FIs,enabling them to predict and curb risks before they manifest into poten

242、tially reputation-damaging events.This move from reactive detection to proactive prevention will mean better outcomes for consumers and the FIs themselves,avoiding the cumbersome process of tracking down and trying to recover funds after they have already undergone various sequences of cross-border

243、transactions.Moreover,with growing goodwill and interest from boards and executive leadership in investing in AI and GenAI programmes,now is the time for compliance teams to present the case for scalable AI solutions that address todays risks and position the FI to proactively combat future threats.

244、AI-driven solutions offer not just operational efficiency but significant effectiveness improvements in detecting and preventing complex financial crimes,which represent significant competitive advantages in todays environment.The shift from traditional methods to AI-driven solutions is no longer op

245、tional it is essential for the future of financial crime compliance.To capitalise on the full potential of AI,FIs must be proactive,strategic,and forward-thinking.Final Thoughts&Strategic RecommendationsAI-driven solutions offer not just operational efficiency but significant effectiveness improveme

246、nts in detecting and preventing complex financial crimes,which represent significant competitive advantages in todays environment.“33Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends 1.FIs should begin with small-scale AI pilots,without comm

247、itting significant resources,conducting low-impact tests to understand the technologys capabilities and limitations against concise criteria for measuring success.These deployments can be gradually scaled up,ensuring that each stage builds on the successes and learnings of the previous one,including

248、 to address data quality issues and challenges associated with legacy system integration.2.FIs should engage openly with regulators to demonstrate the progress and effectiveness of their AI initiatives,while also showcasing the benefits and risk management capabilities of these initiatives to build

249、regulatory confidence in the technology.Technology providers should be expected to partner with firms to engage with regulators to explain AI design and ethics principles.Overall,engagement with regulators should be framed to help shape guidance for the wider regulated population and promote proacti

250、ve and bold steps to drive these outcomes.Strategic Recommendations:Final Thoughts&Strategic RecommendationsThe recommendations set out below aim to address the key takeaways from the research and provide a roadmap for effective AI adoption and integration within FIs.34Untapped Potential:AI-enabled

251、Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends Final Thoughts&Strategic RecommendationsWith these factors in mind,FIs need to prioritise proactively scaling their AI and GenAI capabilities to stay ahead of the curve.FIs that can position AI as a critical solution and

252、secure board-level commitments to sustained investment in AI technologies will not only build long-term resilience and build competitive advantage,but also demonstrate their leadership in helping to define the future of financial crime compliance.3.FIs should from an early stage start with goals of

253、improving operational efficiency,and then define their next set of goals to include reinvesting some efficiency gains into model risk governance and financial crime risk management.In practical terms,FIs should use the time and resources saved to focus on enhancing controls to prevent financial crim

254、e and look across case work to better understand the vulnerabilities being exploited by criminal actors and use those findings to improve AI models.4.FIs should consider using AI technology to enhance data quality and governance practices,including to address common challenges with data lineage and

255、streamline processes associated with data collection,cleansing and transformation.This will reduce the size of investment in data management,while promoting better governance practices,and freeing up time and resources to plan and execute digital transformation initiatives to improve financial crime

256、 prevention capabilities.5.FIs should implement robust governance and accountability frameworks to support AI adoption and develop clear metrics to measure the success of AI initiatives.This includes ensuring boards and senior management engage in and understand the value and benefits of AI in compl

257、iance functions and how risks are managed,which will help to secure strong sponsorship from leadership.An additional outcome of the top-line sponsorship is enhanced engagement with regulators to promote guidance and support for AI adoption in financial crime risk management.35Untapped Potential:AI-e

258、nabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends About SymphonyAIAbout Regulation AsiaRegulation Asia is the leading source for actionable regulatory intelligence for APAC markets.Through our news,events,and research,we serve over 30,000 financial services profes

259、sionals from more than 1,000 financial institutions,as well as regulatory bodies,exchanges,technology firms and other service providers.Learn more at: https:/ us at:SymphonyAI is a leading enterprise AI SaaS company for digital transformation across the most critical and resilient growth verticals,i

260、ncluding retail,consumer packaged goods,financial services,manufacturing,media,and IT/enterprise service management.SymphonyAI verticals have many leading enterprises as clients.Since its founding in 2017,SymphonyAI has grown rapidly to 3,000 talented leaders,data scientists,and other professionals.

261、A 2024 Microsoft Partner of the Year for Business Transformation AI Innovation,SymphonyAI is a SAIGroup company,backed by a$1 billion commitment from successful entrepreneur and philanthropist Dr.Romesh Wadhwani.Learn more at: https:/ us at:Craig RobertsonFinancial Crime SME APACFinancial ServicesSymphonyAIKim YuHead of Marketing APACFinancial ServicesSymphonyAIBradley MacleanCo-founder&Editorial DirectorRegulation ACopyright 2024 inAsia Media Pte Ltd.All rights reserved.36Untapped Potential:AI-enabled Financial Crime Compliance Transformation in Asia Maturity,Applications&Trends