Old Laws for New Machines

Source details

Opening vignette: “It’s 7:45 a.m. in a central bank’s operations room.” AI monitors liquidity conditions and can flag rapid changes (e.g., a flashing red dot beside a bank whose payment activity has suddenly dropped).
Central claim: AI is now “the silent partner of global finance”—embedded in credit scoring, fraud detection, algorithmic trading, and monetary analysis—bringing both “dazzling” promise (faster insights, earlier warnings, more efficient markets) and large perils (system-wide dynamics, feedback loops, and new fragility).
Conceptual anchor: Reinterprets Isaac Asimov’s robot principles (three laws and the “Zeroth Law”) into an ethic for AI in finance:
- First priority: do no harm to financial stability or consumer integrity.
- Second priority: obey institutional mandates without compromising fairness or safety.
- Third priority: preserve resilience without escaping accountability.
- Overarching: serve the higher goal of maintaining trust in the financial system.

AI failures and bias:
- AI can misinterpret patterns, inherit historical biases, or reinforce market signals that accelerate volatility.
- Hard-coded constraints are fragile; generative AI makes it easy to circumvent broadly accepted principles.
Explainability and institutional responsibility:
- “Explainability—because when an algorithm cannot justify its decision, neither can the institution that depends on it.”
- Current AI models “cannot explain their reasoning; they can generate post hoc narratives,” making explainability an institutional—not a technical—requirement.
Macro-financial amplification:
- AI’s machine-speed reactions can create sharper and more synchronized price adjustments and liquidity shifts than human-driven markets.
- Historical reference: the 2010 flash crash demonstrates how automated systems can impact markets before policymakers can intervene.
- AI reshapes feedback loops and influences credit pricing, asset allocation, and risk premiums in mutually reinforcing ways, which can affect shock transmission and central bank policy response.

Optimization vs. moral trade-offs:
- AI “optimizes whatever goal it is given—profit, accuracy, compliance”—but blind obedience can produce morally harmful outcomes (e.g., excluding high-risk groups).
Role of AI as tool, not decider:
- AI should inform human judgment, not replace it—especially for monetary policy, financial supervision, and crisis management.
- Models trained on past data cannot reliably predict novel shocks; humans and machines must “learn together—each aware of the other’s blind spots.”
Capacity-building imperative:
- Financial actors must cultivate both data literacy and ethical literacy; governance of AI remains a moral task.

Technical resilience:
- Redundancy, monitoring, and testing under extreme scenarios are necessary.
Supervisory access and transparency:
- Regulators should be able to audit AI decisions even when proprietary code is involved; “If an algorithm affects financial stability, it should be open to supervisory scrutiny.”
- Secrecy and black-box models increase systemic fragility; the 2008 global financial crisis is invoked as a parallel of complexity without transparency.
Institutional governance:
- Financial institutions should have AI risk officers analogous to chief risk or compliance officers to ensure explainability and auditability.
- Regulators must develop AI literacy and tools to validate and challenge firms’ AI models.
Emerging supervisory tools:
- The Bank for International Settlements (BIS) Innovation Hub has developed prototype tools to help supervisors analyze large datasets and detect anomalies.

Positive use cases:
- Kenya: mobile-money ecosystems such as M-Pesa generate transaction footprints analyzed by AI-based scoring models that expand credit access for small entrepreneurs and previously unbanked populations.
- India: digital identity systems and real-time platforms paired with machine learning aim to better target government transfers and expand microloan access.
Risks of data poverty:
- Limited or biased data can render communities invisible to algorithms; underrepresentation of women, rural populations, or informal workers can entrench exclusion.
International support:
- “The IMF and the World Bank are increasingly integrating digital finance and AI governance issues into their capacity-building programs.”

Regulatory speed and cross-border challenges:
- “AI moves faster than regulation and across borders faster than money.” Cross-border coordination is essential to prevent fragmentation.
Possible international framework:
- Proposes a global set of principles—analogous to the Basel Core Principles for banking—emphasizing fairness, explainability, accountability, and proportionality, allowing consistency with flexibility.
Roles of international institutions:
- IMF: use surveillance and technical assistance to help countries identify AI-related financial risks, share best practices, avoid a digital divide in supervision, and attract skilled professionals from research and fintech communities.
- BIS: could host a repository of supervisory algorithms for collaborative open-source models.
- World Bank and regional development institutions: build AI capacity and digital infrastructure in emerging markets, strengthen data governance, and integrate ethical AI standards into digital finance ecosystems.
Strategic aim: digital multilateralism to ensure AI benefits are global and no country manages these dynamics alone.

Core normative priorities distilled from Asimov’s laws:
- Safety before obedience.
- Transparency before secrecy.
- Trust before profit.
Concrete governance prescriptions (policy-oriented bullets):
- Rigorous testing, sound data governance, and institutional explainability obligations.
- Maintain human discretion in monetary policy, supervision, and crisis management; AI should inform, not decide.
- Build technical resilience: redundancy, monitoring, and extreme-scenario testing.
- Ensure supervisory access to algorithmic decisions; develop supervisory tools and AI literacy.
- Create institutional roles: AI risk officers in financial institutions.
- Coordinate internationally: develop global principles and repositories for supervisory algorithms; expand capacity-building in emerging markets via IMF, World Bank, BIS, and regional institutions.
Overarching message: With intelligent governance grounded in human-centered principles, “AI could become a stabilizing force rather than a source of fragility,” extending inclusion, enhancing oversight, and strengthening monetary legitimacy.

International Monetary Fund

Old Laws for New Machines
- Old Laws for New Machines (Markdown version)
- Old Laws for New Machines (PDF)