Training Programs

Responsible AI for Capital Markets: Surveillance, Compliance, Conduct and Control

 Training positioning:

This programme is designed for AFCM member institutions that operate, regulate, supervise, intermediate or support capital markets. It therefore addresses both market infrastructure perspectives and broker/dealer compliance, conduct, supervisory and operational-risk perspectives. Agentic AI use cases are included as a distinct theme, with appropriate control boundaries.

 Overall Learning Objectives:

• Understand practical AI and agentic AI applications in market surveillance, compliance, investigations and broker/dealer supervision.

• Assess where AI can reduce false positives, detect complex patterns, support case investigation and improve regulatory reporting.

• Identify AI-specific risks, including model risk, explainability limits, hallucination, bias, data leakage, cyber risk and over-reliance.

• Design practical AI governance arrangements aligned with capital markets expectations, three-lines-of-defense models and responsible AI principles.

• Apply an AI risk assessment template to real surveillance and broker/dealer compliance use cases.

• Develop a realistic implementation roadmap from pilot to controlled production deployment.

 

 Day 1 - Detailed Agenda

 Day 1 theme: AI for market surveillance, compliance, enforcement support and broker/dealer supervision.

 1. Opening & Capital Markets Context

 • Why AI matters for AFCM markets now

• Exchange/regulator/CSD perspectives versus broker/dealer perspectives

• From rules-based alerts to AI-assisted supervisory intelligence

• Key opportunity areas: market abuse, AML, communications surveillance, regulatory reporting, conduct risk and internal controls

 

2. AI Fundamentals for Surveillance & Compliance Professionals

 

• AI, machine learning, generative AI and agentic AI: practical distinctions

• Supervised, unsupervised and semi-supervised learning

• Anomaly detection, pattern recognition and predictive analytics

• Natural language processing for investigations, policies, alerts and communications

• Why AI supports, but does not replace, supervisory judgment

 

3. Core Use Cases: AI in Market Surveillance

 

• Detection of abnormal trading patterns and outlier behavior

• Spoofing, layering, wash trading and marking-the-close indicators

• Insider trading, information leakage and pre-announcement activity

• Pump-and-dump, coordinated activity and social-media driven manipulation

• Cross-market, cross-asset, cross-broker and cross-account correlation

• Alert prioritization and false-positive reduction

 

4. Broker/Dealer Compliance and Conduct Supervision

 

• AI for trade surveillance within broker/dealers and securities firms

• Suitability, best execution, churning, unauthorized trading and sales-practice risk indicators

• Client risk profiling, beneficial ownership signals and KYC refresh prioritization

• AML transaction monitoring, sanctions screening and suspicious activity triage

• Communications surveillance across email, chat, recorded calls and collaboration platforms

• Supervisory controls for branches, relationship managers, traders and sales teams

 

5. Agentic AI Use Cases in Surveillance & Compliance

 

• AI investigation copilot: summarizes alerts, drafts case chronologies and identifies missing evidence

• Regulatory reporting agent: prepares first drafts of recurring reports and exception explanations

• Policy and rule-mapping agent: maps new regulatory obligations to internal procedures and controls

• Member/broker supervision agent: compiles broker risk profiles from surveillance, compliance and operational data

• Evidence-gathering workflow agent: requests data, checks completeness, builds timelines and flags inconsistencies

• Control boundary: agentic AI should recommend, draft and route — not make enforcement decisions or execute irreversible actions without human approval

 

6. Data Foundations for Effective Surveillance AI

 

• Order book, trade, client, account, beneficial ownership, communications and market news data

• Data quality, lineage, timestamps, entity resolution and audit trails

• Local language, Arabic/English/Turkish/French data and regional market-structure challenges

• Confidentiality, privacy and cross-border data-transfer constraints

 

7. Interactive Case Study: Market Abuse and Broker Linkages

 

• Scenario: unusual price rise before a material announcement across multiple brokers and related accounts

• What rule-based systems detect versus what AI may add

• How to separate suspicious indicators from legitimate trading activity

• When to escalate, what evidence is needed and what human review must confirm

 

8. Day 1 Wrap-Up

 

• AI is strongest when it improves prioritization, correlation, investigation speed and control coverage

• Human accountability remains central for judgment, escalation and enforcement

 

 

Day 2 - Detailed Agenda

 

Day 2 theme: AI governance, risk management and responsible AI for exchanges, regulators, CSDs and broker/dealers.

 

1. Recap and Governance Transition

 

• Review of Day 1 use cases and where governance must be embedded

• AI risk as legal, conduct, operational, technology, cyber, model, reputational and regulatory risk

 

2. AI Risk Taxonomy for Capital Markets Institutions

 

• Model risk, data risk, bias, drift, explainability and validation limits

• False positives, false negatives and enforcement-quality evidence risk

• Generative AI hallucination, prompt-injection and confidential-data leakage

• Agentic AI risks: tool misuse, unauthorized actions, escalation failure and automated error propagation

• Vendor, outsourcing, cyber, resilience and business-continuity risks

• Special broker/dealer risks: client harm, unsuitable advice, unfair treatment, unauthorized trading and sales-practice issues

 

3. Responsible AI Principles

 

• Human accountability and clear decision ownership

• Transparency, explainability and contestability

• Fairness, non-discrimination and proportionality

• Data protection, confidentiality and lawful use

• Robustness, accuracy, resilience and auditability

• Human-in-the-loop, human-on-the-loop and human-in-command control models

 

 

4. AI Governance Framework for Exchanges, Regulators and Broker/Dealers

 

• Board and senior management accountability for AI adoption

• AI use-case inventory and risk classification

• Three lines of defense: business ownership, independent risk/compliance oversight and internal audit assurance

• Model approval, testing, validation, monitoring and periodic review

• Data governance, access control, retention and lineage

• Agentic AI approval framework: permitted tools, action limits, logs, escalation rules and kill-switches

• Vendor due diligence, service-level commitments, audit rights and exit planning

• Incident management, breach notification and regulatory engagement

 

5. Global Regulatory Landscape and Standards

 

• IOSCO perspectives on AI, market intermediaries and market conduct

• OECD AI Principles and risk-based governance

• EU AI Act relevance for financial services and high-risk systems

• NIST AI Risk Management Framework

• ISO/IEC 42001 AI Management System

• Model-risk-management references and implications for capital market institutions

• MENA/AFCM regional considerations: market maturity, language, data availability and supervisory capacity

 

6. Practical AI Risk Assessment Exercise

 

• Scenario: an exchange or broker/dealer implements an AI system to prioritize suspicious alerts and generate preliminary investigation summaries

• Participants assess purpose, data inputs, key risks, human oversight, controls, documentation and monitoring

• Outputs: short AI risk assessment summary and minimum governance checklist

 

7. Implementation Roadmap for AFCM Members

 

• Start with AI readiness, data readiness and governance readiness

• Prioritize high-value, controlled-risk use cases

• Build AI literacy for surveillance, compliance, legal, risk and technology teams

• Pilot before production; measure false-positive reduction, investigation time saved and control effectiveness

• Scale responsibly through governance gates, model monitoring and independent assurance

 

8. Closing Discussion and Key Takeaways

 

• AI can enhance market integrity, supervisory effectiveness and compliance efficiency

• Agentic AI creates powerful workflow opportunities but requires stronger guardrails

• Responsible adoption requires governance before scale

 

Download the Brochure

 

For registration and inquiries:

Sally Yassin
Sally.yassin@arab-exchanges.org