L1
·
Quiz
·
Lab
L2
·
Quiz
·
Lab
L3
·
Quiz
·
Lab
L4
·
Quiz
·
Lab
Module Test
Module 6 · Lesson 1

Central Bank Digital Currencies

Governments reinvent money itself — and AI decides who gets to spend it.
When a central bank issues a digital currency, what changes — and what risks emerge?

In April 2021, China distributed 40 million yuan in digital currency to citizens via lottery — money that could expire if unspent. The e-CNY pilot, running in Shenzhen, Suzhou, and Chengdu, was the world's largest CBDC experiment to date. Every transaction was logged, traceable, and programmable. The People's Bank of China had not merely digitized cash; it had created a financial instrument that behaves like software.

What Is a CBDC?

A Central Bank Digital Currency is a digital liability of a central bank — the same legal status as a paper banknote, but existing as code. Unlike bank deposits, CBDCs are direct claims on the government. Unlike Bitcoin, they are not decentralized; a central authority controls issuance, supply, and rules.

As of 2024, over 130 countries representing 98% of global GDP are exploring CBDCs, according to the Atlantic Council CBDC Tracker. Three are fully launched: the Bahamas Sand Dollar (2020), the Eastern Caribbean DCash (2021), and China's e-CNY (2022 expansion to 26 cities).

The AI Layer

CBDCs become dramatically more powerful — and more dangerous — when AI is added. Because every transaction is digital from birth, AI can analyze spending patterns in real time, flag suspicious flows instantly, and enforce rules programmatically. The e-CNY wallet app uses machine learning to detect fraud and anomalous transfers.

But programmability cuts both ways. A CBDC can be coded to expire (forcing spending), restricted to certain merchants (controlling what you buy), or frozen for individuals on watchlists — all without a court order, processed by algorithm. This is not speculation: China's social credit system already intersects with financial access.

Real Case — Nigeria eNaira (2021)

Nigeria launched the eNaira in October 2021. Adoption was poor: within a year, fewer than 0.5% of Nigerians used it. The government responded in 2023 by restricting ATM cash withdrawals to force digital adoption — triggering protests and a legal challenge. Compulsory digitization revealed the tension between policy ambitions and public trust.

Key Architecture Choices
Retail vs. Wholesale

Retail CBDCs are held directly by citizens (e-CNY model). Wholesale CBDCs are used only between banks (Project mBridge, BIS). Retail versions raise surveillance concerns; wholesale versions mainly modernize settlement.

Account-Based vs. Token-Based

Account-based systems tie currency to an identity (full traceability). Token-based systems, like physical cash, can be transferred without linking to identity. The EU's digital euro is designed to be privacy-preserving for small amounts.

Programmability

Smart contracts can embed rules: expiry dates, merchant restrictions, automatic tax withholding. Sweden's Riksbank e-krona pilot tested programmable payments for government transfers.

Offline Capability

The Bahamas Sand Dollar supports offline transactions via NFC cards — critical for hurricane-prone islands where internet connectivity fails. This is the primary design rationale for the world's first launched CBDC.

The Privacy Debate

In 2022, the European Central Bank published research showing that 80% of eurozone citizens rank privacy as their top concern about a digital euro. In response, the ECB proposed a tiered system: small transactions below a threshold would remain anonymous; larger flows would require identity verification.

The U.S. Federal Reserve has been more cautious. In 2022, Chair Jerome Powell stated the Fed would not issue a CBDC without explicit authorization from Congress, citing the need for a "broad social consensus." A 2023 executive order from the White House directed agencies to study risks but stopped short of committing to a U.S. digital dollar.

Key Terms
CBDCCentral Bank Digital Currency — digital money issued directly by a central bank as a liability, equivalent in legal status to cash.
Programmable MoneyCurrency with embedded rules enforced by code — expiry, spending restrictions, automatic deductions.
Financial SurveillanceThe ability of a currency issuer to monitor all transactions in real time, enabled by full digitization.

Lesson 1 Quiz

Central Bank Digital Currencies
1. Which country launched the world's first operational CBDC in 2020?
Correct. The Bahamas Sand Dollar launched in October 2020, making it the first fully deployed CBDC. Its primary design rationale was providing financial access in hurricane-prone islands with unreliable internet.
Not quite. China's e-CNY expanded significantly in 2022, but the Bahamas Sand Dollar was the world's first live CBDC, launched in 2020.
2. What made China's 2021 e-CNY lottery unusual compared to traditional cash handouts?
Correct. The e-CNY distributed in China's lottery had programmable expiry dates — an unprecedented feature demonstrating how digital money can be engineered to force economic behavior.
Incorrect. The defining feature of the e-CNY pilot was programmable expiry — the digital yuan could expire if recipients didn't spend it within a set window.
3. Why did Nigeria's eNaira pilot face such low adoption in its first year?
Correct. The eNaira's first year saw less than 0.5% adoption despite official promotion, reflecting deep public distrust. The government eventually resorted to restricting cash withdrawals to force digital use — triggering protests.
Incorrect. The issue was voluntary adoption failure — fewer than 0.5% of Nigerians used the eNaira in its first year, leading to coercive restrictions that sparked protests.
4. What concern did 80% of eurozone citizens rank highest about a potential digital euro?
Correct. ECB research in 2022 found privacy was the dominant concern among eurozone citizens, prompting proposals for tiered anonymity: small transactions stay private, larger ones require identity verification.
Not right. The ECB's 2022 research showed privacy as the overwhelming top concern — prompting the ECB to design a tiered system where small transactions would remain anonymous.

Lab 1 — CBDC Design Advisor

Explore the policy trade-offs in designing a central bank digital currency.

Your Scenario

You are advising a fictional small island nation on whether and how to launch a CBDC. You must weigh financial inclusion, privacy rights, technical feasibility, and government control. Use this AI advisor to think through the design choices and trade-offs.

Starter prompts: "What are the biggest risks of a retail CBDC for a country with low smartphone penetration?" — or — "How should we handle privacy if citizens don't trust the government?" — or — "Compare the Bahamas Sand Dollar approach to China's e-CNY."
CBDC Policy Advisor
AI Lab
Welcome. I'm your CBDC design consultant. You're advising a small island nation on launching a central bank digital currency. What aspects of design or policy would you like to explore first — privacy architecture, financial inclusion, programmability, or the geopolitical implications of your currency choice?
Module 6 · Lesson 2

DeFi, Stablecoins, and AI-Managed Portfolios

Algorithms replace brokers. Code replaces contracts. Who is responsible when things go wrong?
Can AI and decentralized finance displace traditional financial intermediaries — and at what cost?

In May 2022, the Terra/Luna ecosystem — a $60 billion algorithmic stablecoin system — collapsed in 72 hours. TerraUSD (UST) was meant to hold its $1 peg through an algorithmic relationship with Luna tokens, governed by smart contracts with no human override. When a coordinated sell-off began on May 7, the algorithm's automatic responses accelerated the death spiral rather than halting it. By May 13, Luna had lost 99.9% of its value. An estimated $40 billion in wealth was destroyed. No regulator, no court, no algorithm saved it.

What Is DeFi?

Decentralized Finance refers to financial services — lending, borrowing, trading, earning yield — built on public blockchains using smart contracts, without banks or brokers. By peak 2021, over $180 billion in assets were locked in DeFi protocols. The largest, Aave and Compound, allowed users to lend and borrow cryptocurrency algorithmically, with interest rates set by code based on supply and demand.

AI enters DeFi in several ways: automated market makers (AMMs) use mathematical formulas to set prices without order books; liquidation bots monitor collateral ratios and automatically sell positions when thresholds are breached; and MEV (maximal extractable value) bots use AI to front-run transactions for profit.

Stablecoins: Three Architectures
Fiat-Backed

Each token backed 1:1 by dollars in a bank. Tether (USDT) and USD Coin (USDC) dominate. Tether's reserves were found partly in commercial paper (2021 CFTC settlement, $41M fine). Circle's USDC briefly depegged in March 2023 when $3.3B was held at Silicon Valley Bank during its collapse.

Crypto-Collateralized

MakerDAO's DAI is backed by crypto assets (over-collateralized to absorb volatility). An AI-adjacent system of smart contracts monitors collateral ratios and auto-liquidates undercollateralized positions. DAI survived the 2022 crash, validating over-collateralization as a stabilizing design.

Algorithmic

No collateral — purely algorithmic stabilization. Terra/Luna was the largest example and proved catastrophically fragile. The algorithm was not robust to coordinated attacks. Post-Luna, algorithmic stablecoins have largely lost market credibility.

Real Case — Three Arrows Capital, 2022

Three Arrows Capital (3AC), once a $10 billion crypto hedge fund, collapsed in June 2022 after automated DeFi liquidations wiped out leveraged positions in Luna and stETH simultaneously. 3AC had borrowed billions from centralized lenders (Celsius, Voyager, BlockFi) without adequate disclosure. When on-chain liquidation bots automatically closed 3AC's positions, it triggered a chain of insolvencies across the crypto lending ecosystem — a domino effect no algorithm was designed to prevent.

AI-Managed Portfolios in Traditional Finance

Outside crypto, AI portfolio management has become mainstream. Betterment and Wealthfront, founded in 2008–2011, pioneered robo-advisors — algorithms that auto-allocate assets based on risk tolerance and rebalance portfolios automatically. By 2023, robo-advisors managed over $1 trillion in assets globally.

BlackRock's Aladdin platform uses machine learning to manage risk across $21.6 trillion in assets (as of 2022) — monitoring portfolios for over 4,000 asset managers. Aladdin runs 5,000 portfolio stress tests daily. Regulators have noted that because Aladdin is so widely used, correlated behavior during a crisis could itself become a source of systemic risk.

The Correlated Risk Problem

When many institutions use similar AI models, they may make the same trades simultaneously — amplifying market moves rather than dampening them. The 2010 Flash Crash, in which the Dow dropped 1,000 points in minutes before recovering, involved algorithmic trading feedback loops. The SEC and CFTC's joint report identified automated systems as the primary mechanism of the crash.

DeFiDecentralized Finance — financial services built on blockchain smart contracts without traditional intermediaries.
AMMAutomated Market Maker — an algorithm that sets asset prices based on a mathematical formula rather than a traditional order book.
Algorithmic StablecoinA cryptocurrency designed to maintain price stability through algorithmic supply adjustments rather than collateral backing.

Lesson 2 Quiz

DeFi, Stablecoins, and AI-Managed Portfolios
1. How much wealth was estimated to have been destroyed in the Terra/Luna collapse of May 2022?
Correct. Approximately $40 billion in wealth was destroyed in 72 hours as the algorithmic stabilization mechanism failed under coordinated selling pressure — with no human or algorithmic override available.
Incorrect. The Terra/Luna collapse destroyed an estimated $40 billion in 72 hours — one of the fastest and largest financial collapses in history.
2. Which stablecoin type survived the 2022 crypto crash, validating its design approach?
Correct. DAI's over-collateralized design meant it had enough buffer to withstand the crypto crash. Its smart contract liquidation system automatically closed undercollateralized positions before they threatened the peg.
Incorrect. DAI (MakerDAO) survived the crash by being over-collateralized with crypto assets, automatically liquidating risky positions via smart contracts before they could threaten the peg.
3. What is a key systemic risk identified by regulators regarding BlackRock's Aladdin platform?
Correct. Regulators have noted that Aladdin's dominance — overseeing $21.6 trillion in assets — means correlated algorithmic behavior during a crisis could amplify rather than dampen market stress.
Incorrect. The identified risk is correlated behavior — when so many institutions use the same AI system, they may all execute similar trades simultaneously, turning a small shock into a crisis.
4. What role did automated liquidation bots play in the collapse of Three Arrows Capital?
Correct. On-chain liquidation bots mechanically closed 3AC's undercollateralized positions, triggering cascading insolvencies at Celsius, Voyager, and BlockFi — none of which had adequate disclosure about their 3AC exposure.
Incorrect. DeFi liquidation bots automatically closed 3AC's leveraged positions when collateral ratios fell below thresholds, setting off a chain reaction that bankrupted multiple centralized lenders.

Lab 2 — DeFi Risk Analyst

Diagnose systemic risks in decentralized finance and algorithmic trading systems.

Your Scenario

You are a risk analyst at a financial stability board tasked with evaluating DeFi protocols and AI-managed investment systems for systemic vulnerabilities. Use this AI analyst to probe the failure modes of specific systems and propose regulatory safeguards.

Starter prompts: "What mechanisms caused Terra/Luna to collapse so rapidly?" — or — "How should regulators treat automated liquidation bots that accelerate market crashes?" — or — "What's the difference between a robo-advisor and a DeFi AMM in terms of who bears risk?"
DeFi Risk Analyst
AI Lab
I'm your DeFi and algorithmic finance risk analyst. We can examine stablecoin architectures, liquidation cascades, correlated AI trading risks, or specific protocol failures. What aspect of decentralized or algorithmic finance would you like to stress-test today?
Module 6 · Lesson 3

AI in Payments and Financial Inclusion

Two billion unbanked people. One mobile phone. The question is whether AI serves them or extracts from them.
Can AI-powered payments reach people that traditional banks left behind — and what new risks does this create?

When Safaricom launched M-Pesa in Kenya in 2007, banks dismissed it as a telecom product. Within five years, it was processing more transactions than Western Union did globally. By 2023, M-Pesa had 51 million customers across seven African countries and processed over $314 billion in transactions annually. It was built not on AI but on simple SMS — yet it proved that mobile money could reach the financially excluded at scale. Today, AI layers on top of platforms like M-Pesa are extending credit, flagging fraud, and setting interest rates for borrowers with no credit history.

The Scale of Financial Exclusion

According to the World Bank's Global Findex 2021, 1.4 billion adults remain unbanked — without any account at a financial institution or mobile money provider. They are disproportionately women (56% of the unbanked), rural residents, and people in low-income countries. The barriers: no documentation, no nearby branch, minimum balance requirements, distrust of institutions.

Mobile money has made the largest dent. Between 2011 and 2021, the share of adults with a financial account rose from 51% to 76% globally — mostly driven by mobile money accounts in Sub-Saharan Africa. AI is now the next layer: making lending decisions, detecting fraud, and personalizing financial products for people with thin or no credit files.

AI Credit Scoring for the Unbanked

Tala (formerly InVenture), founded in 2014, was among the first fintechs to use smartphone behavioral data — app usage patterns, call logs, location data, spending patterns — to make microloan decisions in Kenya, Tanzania, India, and the Philippines. By 2022, Tala had disbursed over $3 billion in loans to 7 million customers, most of whom had never held a bank account.

Similarly, Branch International uses machine learning on mobile data to assess creditworthiness. Its models analyze over 10,000 data points per applicant — including the battery charge level when you apply (lower charge correlates with lower repayment, per their research). By 2023, Branch had disbursed $700 million in loans across Africa and India.

Risk — Predatory Lending at Scale

In Kenya, the rapid proliferation of mobile lending apps — over 110 registered by 2019 — led to a debt crisis among low-income borrowers. A 2019 Financial Sector Deepening Kenya study found that 1 in 5 mobile loan users had defaulted, and many had taken loans from multiple apps simultaneously. Interest rates on 30-day loans commonly exceeded 100% APR. Kenya's Central Bank responded with the Digital Credit Providers Regulation in 2022, requiring registration and rate disclosure — the first such regulation in Africa specifically targeting AI-based mobile lenders.

Real-Time Payments Infrastructure

Traditional bank transfers in the U.S. took 1–3 days until the Federal Reserve launched FedNow in July 2023 — a real-time payment system that settles transfers in seconds, 24/7. India's Unified Payments Interface (UPI), launched in 2016, is the most successful example globally: by 2023, UPI processed over 10 billion transactions per month, accounting for over 46% of global real-time payment volume.

AI is embedded throughout UPI's fraud detection layer. The National Payments Corporation of India (NPCI) uses machine learning models that analyze transaction patterns in real time to block fraudulent payments before settlement. In FY2023, UPI's fraud rate was approximately 0.0015% by value — one of the lowest for any payment system at scale.

2007
M-Pesa launches in Kenya — first mass-market mobile money, SMS-based. Proves financial inclusion via mobile is possible.
2016
India UPI launches — real-time bank transfers via mobile. Reaches 10B+ transactions/month by 2023.
2018
WeChat Pay / Alipay — combined serve over 1.5 billion users in China. AI fraud detection processes millions of transactions per second.
2022
Kenya Digital Credit Regulation — first African law specifically targeting algorithmic mobile lenders, requiring registration and APR disclosure.
2023
FedNow launches in the U.S. — real-time settlement for the first time. AI fraud monitoring embedded at the infrastructure level.
Key Terms
Thin-File BorrowerA person with little or no credit history, making traditional credit scoring impossible. AI alternative data approaches attempt to serve this population.
Alternative DataNon-traditional data used for credit decisions: mobile usage patterns, location history, app behavior, social connections.
Real-Time PaymentsPayment infrastructure that settles transactions in seconds rather than days, operating continuously. UPI and FedNow are leading examples.

Lesson 3 Quiz

AI in Payments and Financial Inclusion
1. How many adults worldwide remained unbanked according to the World Bank's Global Findex 2021?
Correct. 1.4 billion adults remained unbanked as of 2021 — disproportionately women, rural residents, and people in low-income countries. This is the population AI-powered fintech aims to serve.
Incorrect. The World Bank found 1.4 billion unbanked adults globally in 2021, a significant decline from 2011 but still representing a major gap in financial access.
2. What unusual data point does Branch International's AI model reportedly use as a credit signal?
Correct. Branch International found that lower battery charge at application time correlates with lower repayment rates — one of over 10,000 behavioral signals its models analyze. This illustrates how alternative data can be predictive but also raises fairness questions.
Incorrect. Branch International's research found that battery level at application time is actually a predictive signal — lower charge correlates with lower repayment. It's one of 10,000+ behavioral data points their models analyze.
3. What fraction of global real-time payment volume did India's UPI account for in 2023?
Correct. India's UPI accounted for over 46% of global real-time payment volume in 2023, processing more than 10 billion transactions per month — making it the world's dominant real-time payments system.
Incorrect. India's UPI dominates global real-time payments, accounting for over 46% of volume in 2023 with 10+ billion monthly transactions — driven by near-universal smartphone adoption and government mandates.
4. What triggered Kenya's 2022 Digital Credit Providers Regulation?
Correct. Kenya's regulation responded to a debt crisis: over 110 mobile lending apps had proliferated by 2019, many charging 100%+ APR, with 1 in 5 borrowers defaulting. The regulation became the first in Africa to specifically target AI-based mobile lenders.
Incorrect. The regulation was triggered by a debt crisis from unregulated mobile lending — 110+ apps offering loans at 100%+ APR with 20% default rates, causing significant harm to low-income Kenyans.

Lab 3 — Financial Inclusion Strategist

Design AI-powered payment and lending systems that expand access without causing harm.

Your Scenario

You are a program officer at a development finance institution funding fintech expansion in East Africa. You need to evaluate which AI-based approaches genuinely expand financial inclusion and which risk repeating Kenya's predatory lending crisis. Use this AI strategist to think through responsible design.

Starter prompts: "What safeguards should an AI credit scoring system have for thin-file borrowers?" — or — "How did India's UPI succeed where other real-time payment systems struggled?" — or — "Is using behavioral data like battery level for credit scoring fair or discriminatory?"
Financial Inclusion Strategist
AI Lab
Welcome. I'm your financial inclusion strategist. We're evaluating AI-powered fintech programs for East Africa — weighing genuine access expansion against risks of predatory lending and data exploitation. What aspect would you like to examine: credit scoring models, payment infrastructure, regulatory design, or consumer protection?
Module 6 · Lesson 4

Regulation, Accountability, and the Next Frontier

Who governs algorithms that govern money? The regulatory frameworks being built now will define financial AI for decades.
As AI assumes more control over financial systems, what governance structures can prevent harm while enabling innovation?

In October 2021, the SEC and CFTC held a joint hearing on AI in financial markets. A commissioner asked the fundamental question: "When an algorithm causes harm — to a retail investor, to market stability — who is accountable?" Representatives from Citadel Securities, Virtu Financial, and several fintech firms gave a version of the same answer: the algorithm reflects human choices, therefore existing laws apply. Regulators were not satisfied. Within a year, the SEC proposed rules requiring broker-dealers to evaluate AI systems for conflicts of interest — the first direct attempt to govern AI financial decision-making in U.S. law.

The Global Regulatory Landscape

Financial AI regulation is developing unevenly across jurisdictions. The EU has the most comprehensive framework: the AI Act (passed 2024) classifies AI systems used in credit scoring and risk assessment as "high risk," requiring transparency, human oversight, and explainability. The EU's Digital Markets Act and Digital Services Act add further constraints on large platform companies offering financial services.

The U.S. approach remains fragmented. The SEC issued guidance in 2023 requiring registered investment advisers to disclose AI use in portfolio management. The Consumer Financial Protection Bureau (CFPB) issued guidance in 2023 requiring lenders using AI credit models to provide specific reasons for denials — not just "the model said no."

Real Case — SEC vs. AI "Conflicts of Interest" (2023)

In July 2023, the SEC proposed a rule requiring broker-dealers and investment advisers to neutralize conflicts of interest embedded in AI and predictive analytics tools — specifically targeting systems that optimize for firm profit rather than client benefit. The proposal targeted the practice of using AI to nudge investors toward higher-fee products. It received 4,600 public comments and remained under review as of 2024, illustrating how contentious AI financial regulation has become.

Key Regulatory Challenges
Explainability

Modern deep learning credit models can be highly accurate but uninterpretable. The CFPB's 2023 guidance confirmed the Equal Credit Opportunity Act requires specific, accurate reasons for denials — "our AI model" is not sufficient. Lenders must be able to explain which factors drove a decision.

Algorithmic Discrimination

In 2021, the U.S. Department of Housing and Urban Development settled with Facebook over its AI ad targeting system, which allowed advertisers to exclude users by race from seeing housing ads. Similar proxy discrimination concerns apply to credit AI using zip codes, device type, or behavioral data as proxies for protected classes.

Speed vs. Oversight

High-frequency trading operates in microseconds — no human can review trades in real time. The SEC's 2023 market structure reforms proposed circuit breakers and minimum resting times for orders to slow algorithmic trading and allow market makers to maintain fair prices.

Cross-Border Gaps

Crypto and DeFi protocols operate globally with no single regulator. The 2022 collapse of FTX — a Bahamas-registered exchange — left U.S. customers with $8 billion in losses and no deposit insurance. The subsequent crypto regulatory push in the U.S. (FIT21 Act, 2024) attempted to clarify jurisdiction between SEC and CFTC.

Emerging Frontiers: What Comes Next

Generative AI in financial advice is the next major frontier. In 2023, Morgan Stanley deployed a GPT-4 powered tool for its 16,000 financial advisors — helping them synthesize research, draft client communications, and find relevant internal data. Goldman Sachs, JPMorgan, and Bloomberg all announced similar AI integration plans. The risk: AI advisors that confidently give wrong information, or that embed subtle biases from training data into financial recommendations.

The concept of programmable money + AI represents perhaps the most profound shift: when your central bank digital currency can automatically enforce tax payments, restrict spending during financial emergencies, or implement monetary policy at the level of individual wallets, the relationship between citizens and financial systems changes fundamentally. This is not dystopian speculation — it is the explicit design goal of several CBDC programs under development.

The Accountability Gap

When an algorithm denies your loan, freezes your account, or executes a trade that loses your savings, the accountability question remains unresolved in most legal systems. Traditional financial liability frameworks assume human decision-makers. Distributed AI systems — where no single actor designed the outcome — create what legal scholars call the "many hands problem." Resolving this may require new legal concepts: algorithmic liability, mandatory explainability, or mandatory human-in-the-loop for consequential financial decisions.

EU AI ActThe world's first comprehensive AI law (2024), classifying credit scoring and risk assessment AI as "high risk" requiring transparency, oversight, and explainability.
Many Hands ProblemThe difficulty of assigning responsibility for AI harms when the outcome results from many distributed decisions by developers, trainers, deployers, and users.
Programmable MoneyCurrency that can automatically enforce rules at the transaction level — tax payments, spending restrictions, expiry dates — without human intervention.

Lesson 4 Quiz

Regulation, Accountability, and the Next Frontier
1. Under the EU AI Act (2024), how are AI systems used for credit scoring classified?
Correct. The EU AI Act classifies credit scoring and financial risk assessment AI as "high risk," imposing requirements for transparency, human oversight, audit trails, and explainability to affected individuals.
Incorrect. The EU AI Act places credit scoring in the "high risk" category — one of the most regulated tiers — requiring transparency, explainability, and meaningful human oversight.
2. What did the CFPB's 2023 guidance require of lenders using AI credit models?
Correct. The CFPB confirmed that the Equal Credit Opportunity Act requires lenders to provide specific, accurate reasons for credit denials — AI opacity is not an acceptable explanation. Lenders must be able to identify which factors drove the denial.
Incorrect. The CFPB's guidance required lenders to explain AI credit denials specifically — "the model declined you" is legally insufficient. Lenders must identify the actual factors that caused the denial.
3. What was the target of the SEC's July 2023 proposed rule on AI in financial services?
Correct. The SEC's 2023 proposal targeted AI-driven conflicts of interest — specifically systems that nudge investors toward higher-fee products in ways that benefit the firm over the client. It received 4,600 public comments, reflecting how contentious this regulation became.
Incorrect. The SEC's 2023 proposal focused on conflicts of interest in AI systems — targeting tools that optimize for firm revenue at the expense of clients, particularly through steering investors toward high-fee products.
4. What is the "many hands problem" in AI financial accountability?
Correct. The many hands problem describes how distributed AI development makes traditional liability concepts hard to apply — when developers, data curators, model trainers, deployers, and users all contributed, no single party is clearly responsible for a harmful outcome.
Incorrect. The many hands problem is a legal and ethical concept: when an AI harm results from contributions by many parties across the development pipeline, traditional liability frameworks — designed for single human decision-makers — break down.

Lab 4 — AI Finance Regulatory Counsel

Navigate the emerging legal and governance frameworks shaping AI in money and markets.

Your Scenario

You are a regulatory counsel advising a fintech company launching an AI-powered investment advisory platform in both the EU and U.S. markets. You must understand compliance requirements, identify potential liability exposure, and design governance structures that satisfy regulators on both sides of the Atlantic.

Starter prompts: "What does the EU AI Act require of our credit scoring model specifically?" — or — "How should we document AI decision-making to satisfy the CFPB's explainability requirements?" — or — "If our AI gives a client bad investment advice, who is liable — us, the model developer, or the cloud provider?"
AI Finance Regulatory Counsel
AI Lab
I'm your AI finance regulatory counsel. We're preparing your EU-U.S. fintech launch for compliance with the AI Act, SEC disclosure rules, and CFPB explainability requirements. Where would you like to start — product classification under the EU AI Act, conflict-of-interest governance for your advisory algorithm, or liability frameworks for AI-driven financial harm?

Module 6 Test

The Future of Money — 15 questions · Pass at 80%
1. Which of the following best describes a CBDC?
Correct. A CBDC is a direct liability of the central bank — not a commercial bank, not a private issuer — and carries the same legal status as physical banknotes.
Incorrect. A CBDC is specifically a direct liability of the central bank — meaning it is backed by the government itself, like physical cash, not by a commercial bank or private entity.
2. The Bahamas Sand Dollar was designed primarily to solve which problem?
Correct. The Sand Dollar's offline NFC capability was the core design requirement — ensuring residents could transact even when internet infrastructure was destroyed by hurricanes.
Incorrect. The Bahamas Sand Dollar's primary design rationale was offline functionality via NFC cards, enabling financial transactions when hurricanes destroy internet infrastructure.
3. What property distinguishes an algorithmic stablecoin from a crypto-collateralized stablecoin like DAI?
Correct. Algorithmic stablecoins rely purely on algorithmic supply and demand adjustments — no collateral backs them. Terra/Luna's failure proved this design is fragile to coordinated attacks.
Incorrect. The key distinction is collateral: algorithmic stablecoins have none — they rely on algorithm-driven token issuance/burning to maintain pegs, which proved catastrophically fragile in Terra/Luna's case.
4. In what year did M-Pesa launch in Kenya, and what technology did it originally use?
Correct. M-Pesa launched in 2007 using simple SMS — proving that financial inclusion at scale did not require smartphones, apps, or AI. It later added AI layers for fraud detection and credit scoring.
Incorrect. M-Pesa launched in Kenya in 2007 using basic SMS technology — accessible on any mobile phone. This simplicity was key to its rapid adoption among low-income users.
5. What was Tala's primary innovation in credit scoring for unbanked borrowers?
Correct. Tala pioneered the use of smartphone behavioral data as a proxy for creditworthiness — enabling loans to 7 million customers across Africa and Asia who had never held bank accounts.
Incorrect. Tala's innovation was using smartphone behavioral data — how you use your phone, your location patterns, your app habits — to make credit decisions for people with no traditional credit history.
6. How much did India's UPI process per month by 2023?
Correct. UPI crossed 10 billion monthly transactions in 2023, accounting for over 46% of global real-time payment volume — making India's payment infrastructure the world's most active.
Incorrect. India's UPI surpassed 10 billion transactions per month in 2023, representing more than 46% of all global real-time payment volume.
7. What specifically caused the 2010 Flash Crash — a 1,000-point intraday Dow drop?
Correct. The SEC/CFTC joint report identified algorithmic trading feedback loops as the primary mechanism — automated systems reacting to each other's behavior in a cascade that briefly erased nearly $1 trillion in market value.
Incorrect. The SEC and CFTC's joint investigation identified algorithmic trading feedback loops as the primary cause — automated systems reacting to each other's behavior in an accelerating cascade.
8. What legal concept captures the difficulty of assigning blame when AI harm results from contributions by many different actors?
Correct. The "many hands problem" describes how traditional liability frameworks, designed for single human decision-makers, break down when developers, trainers, deployers, and users all contributed to a harmful outcome.
Incorrect. Legal scholars call this the "many hands problem" — when so many parties contributed to an AI system's harmful output that no single actor can be clearly assigned responsibility.
9. Which institution manages risk across $21.6 trillion in assets using machine learning — potentially creating systemic risk through correlated behavior?
Correct. BlackRock's Aladdin manages risk for over 4,000 asset managers and runs 5,000 daily stress tests. Its dominance means correlated AI behavior could amplify market stress — a concern regulators have flagged.
Incorrect. BlackRock's Aladdin platform oversees $21.6 trillion in assets across 4,000+ institutional clients. Regulators have noted that its widespread use creates correlated behavior risk.
10. The FTX collapse in 2022 revealed which critical gap in crypto market regulation?
Correct. FTX was registered in the Bahamas, leaving U.S. customers with $8 billion in losses and no deposit insurance, no clear regulator, and no consumer protection — exposing a critical cross-border gap.
Incorrect. FTX's Bahamas registration meant U.S. customers had $8 billion in losses with no deposit insurance and unclear regulatory jurisdiction — highlighting the cross-border gap the FIT21 Act attempted to address.
11. In Nigeria's eNaira experience, what government response to low adoption triggered protests?
Correct. After fewer than 0.5% of Nigerians voluntarily adopted the eNaira, the government restricted cash ATM withdrawals in 2023 — effectively forcing digital use — which sparked widespread protests.
Incorrect. Nigeria restricted ATM cash withdrawals in 2023 to force eNaira adoption, triggering protests — a lesson in the limits of coercive CBDC policy without earned public trust.
12. Which type of DeFi automation is specifically designed to close positions when collateral ratios fall below a threshold?
Correct. Liquidation bots monitor on-chain collateral ratios continuously and automatically sell undercollateralized positions — protecting lenders but capable of triggering cascading liquidations, as seen in the Three Arrows Capital collapse.
Incorrect. Liquidation bots specifically monitor collateral ratios and automatically close positions that fall below the required threshold — they played a key role in the Three Arrows Capital cascade.
13. What was the key finding of Kenya's 2019 FSD Kenya study on mobile lending?
Correct. The 2019 FSD Kenya study found a 20% default rate among mobile loan users, with many trapped in multi-app debt cycles at extremely high interest rates — directly motivating Kenya's 2022 regulatory intervention.
Incorrect. FSD Kenya found a 20% default rate and extensive multi-app borrowing at 100%+ APR — evidence of a predatory lending crisis that ultimately prompted Africa's first regulation specifically targeting algorithmic mobile lenders.
14. What is the primary concern about Morgan Stanley's deployment of GPT-4 for its 16,000 fi