The Market Can No Longer Govern Itself

For decades, economists have debated whether financial markets are efficient, and this question once seemed largely theoretical. Do prices fully reflect available information? Today, that debate has been eclipsed by a more urgent concern. When algorithms set prices, allocate capital, and execute trades at speeds no human can comprehend, who is accountable for their decisions?

The Efficient Market Hypothesis (EMH), formalized by Eugene Fama in 1970, holds that market prices rapidly incorporate all available information. The theory profoundly shaped modern finance, regulation and investment management. Yet it rested on an assumption that no longer holds, namely that human beings were the primary interpreters of information. That is no longer true.

From the invisible hand to the invisible algorithm

Artificial intelligence (AI) has not only sped up financial markets but has also fundamentally changed their structure.

Human traders used to analyze earnings reports, regulatory filings, geopolitical events and macroeconomic indicators amid uncertainty. Now, AI systems process huge amounts of structured and unstructured data, such as satellite images, transaction streams, social media activity, and behavioral signals, in just milliseconds. They detect patterns that surpass human understanding and respond almost immediately.

In principle, faster and more accurate price discovery should enhance market efficiency. The greatest risks, however, arise not from speed itself but from opacity, concentration, and systemic fragility.

As financial institutions increasingly rely on similar AI systems trained on comparable historical data and aligned incentives, markets risk becoming prone to synchronized action. While this convergence might enhance efficiency during normal times, it can escalate instability during crises, leading to coordinated selloffs, liquidity crises and rapid flash crashes that outpace human response capabilities.

This phenomenon, the dominance of a single model, could evolve into one of the key systemic risks in twenty-first-century finance. Having a variety of models is not just a technical choice; it is crucial for resilience.

The boardroom deficit

The implications extend well beyond trading floors.

AI is transforming corporate governance by changing how analytical power is distributed within organizations. Management now has greater control not just over information flows but also over the AI systems that interpret this data. A board that cannot independently scrutinize these systems is not providing genuine oversight; instead, it is only performing symbolic oversight.

This is the Thinking Board problem: the widening gap between boards that simply accept algorithmically curated recommendations and those that can critically analyze the underlying assumptions, incentives and limitations of those recommendations.

Bridging this gap involves more than just technical briefings and executive dashboards. It requires board-level digital literacy, autonomous analytical skills and governance frameworks that treat algorithmic outputs as subjects of scrutiny rather than as authoritative sources.

Without these reforms, achieving intelligence symmetry, the fair distribution of analytical capacity among institutions, will remain out of reach.

When the black box violates the law

Opacity is no longer just a technical issue; it is increasingly a legal and institutional challenge as well.

When an AI system discriminates in lending, enables implicit price coordination, or influences investors to favor products that boost institutional profits over clients’ interests, assigning accountability becomes challenging. Current legal standards usually regard AI as a tool, with responsibility falling on the deploying organizations. However, as AI systems grow more autonomous, adaptable, and integrated into financial infrastructure, it becomes harder to identify who is responsible.

Regulators have started to take action. The US Securities and Exchange Commission has proposed regulations to address conflicts of interest in AI-driven investor interactions. Meanwhile, the European Union’s AI Act designates applications like credit scoring and insurance underwriting as high-risk systems that must meet standards for transparency, testing, and human oversight.

These steps are important, but regulatory bodies still operate on a years-long timeline, whereas the technologies they oversee develop in just months.

Governance is now market infrastructure

Viewing regulation only as a market constraint overlooks the deeper nature of the ongoing transformation.

In AI-driven financial systems, governance has become essential for market efficiency rather than merely an external consideration.

Markets fundamentally depend on trust. When prices are influenced by systems that participants cannot comprehend, audit, or contest, trust erodes. A flash crash is more than just a technical error; it represents a loss of confidence. Similarly, a discriminatory lending algorithm is not merely a compliance concern; it damages the legitimacy of the process.

Effective governance relies on establishing institutional infrastructure capable of managing algorithmic complexity. This includes conducting regular independent audits of AI systems, implementing insurance schemes to evaluate algorithmic risks, creating regulatory sandboxes for stress-testing models prior to crises, and maintaining AI registries that provide regulators with transparency into the deployed systems.

These measures do not hinder innovation; instead, they ensure it is sustainable.

The global governance gap

A rarely addressed aspect of AI-driven finance is that its regulations are primarily being developed in advanced financial hubs in a few countries, leaving much of the Global South on the sidelines.

However, the technologies being regulated will impact financial systems worldwide.

Governance structures created without widespread participation can be not only uneven but also unfair. Standards tailored for advanced economies might unintentionally limit financial inclusion in other regions. Models primarily trained on Global North data may make significant decisions that affect Global South societies without adequately considering their institutional contexts, development goals, or social conditions.

This is why multilateral institutions such as multilateral development banks, the Financial Stability Board, and international standard-setting organizations need to take a more proactive role in developing globally inclusive AI governance frameworks.

The future of financial governance should not be determined by only a few jurisdictions and then imposed on the rest of the world.

What comes next

The Efficient Market Hypothesis was never intended as a literal depiction of reality. Instead, it served as an analytical benchmark for understanding market imperfections. Its lasting importance is in showing where markets do not succeed.

The source of failures has shifted. The main inefficiencies are now more structural than behavioral, including black-box opacity, model monoculture, analytical asymmetries and reliance on complex systems that are rarely fully understood, rather than biases, heuristics and irrationalities as noted by behavioral economics.

Fixing these issues is now a collective responsibility, involving not just economists but also regulators, technologists, lawyers, ethicists, corporate leaders, and governance institutions working at the crossroads of capital and computation.

The key issue in modern finance is no longer whether markets are efficient, but whether the governing institutions can match the intelligence built into those markets.

The invisible hand remains, but it is now more often directed by unseen algorithms.

The effectiveness of these algorithms in serving the public interest hinges not on their technical complexity but on the robustness, legitimacy, and inclusiveness of the institutions that oversee them.