Thursday, November 19, 2026

 

DIGITAL LIFE


AI agents can turn minor issues into financial crises

The stock market is not the economy. Nor is it a direct determinant of financial stability. But when technology stocks plummeted on November 5, hitting exchanges across the globe and erasing trillions of dollars in market valuations, suspicions of a bursting artificial-intelligence bubble – or even a full-fledged crash – did not seem out of place.

Such reactions may now be a natural, new normal. The markets moved on, as they often do – until the Nvidia fallout a week after the post-U.S. Election Day volatility, Both before and since, economists and official watchdogs have been on alert, even issuing warnings of a possible meltdown. That may or may not be preventative, but the scenarios cast light on potentially dire prospects.

At the end of August, Financial Stability Board Chair Andrew Bailey asserted that an already elevated set of economic, credit and geopolitical risks “has been further complicated by the emergence of frontier AI models, which are showing increasingly sophisticated autonomy and problem-solving abilities, as well as threat capabilities.

“This is particularly important from a financial stability perspective,” said Bailey’s letter to the G20 finance ministers and central bankers. “The risks associated with frontier AI will not respect national borders. The global financial system is highly interconnected, and cyber disruption can spread across jurisdictions through common technology providers, shared infrastructure, and cross-border financial activity.

“Differences in legal frameworks, cyber capability, resilience and recovery capacity across jurisdictions could therefore have consequences well beyond the jurisdiction in which an incident originates and may themselves become a source of vulnerability.”

Artificial intelligence is driving billions in investments, transforming companies, and promising to make financial markets more efficient. However, the very technology that excites investors is beginning to cause concern among those responsible for economic stability. The reason goes beyond the valuations of tech giants. Systems capable of making decisions in seconds, moving capital, and reacting automatically to news can create risks that are difficult to control—and a new American survey shows just how much this concern is growing.

On November 13, the Federal Reserve (Fed)—the central bank of the United States—published its semiannual report on financial stability.

Among the findings, one figure stood out: 87% of the participants surveyed identified artificial intelligence as a significant risk to the U.S. financial system.

The survey was conducted between September and October, involving 23 representatives from banks, investment funds, financial intermediaries, and consulting firms.

In the previous survey, conducted in the spring, this concern had been mentioned by 50% of respondents.

This shift reveals a marked increase in risk perception among the participants.

However, an important caveat applies: the 87% figure represents only the specific experts consulted, not all economists or financial market professionals.

Furthermore, the survey does not predict an inevitable crisis.

Concerns include high valuations for AI-related companies, cyberattacks, and the complex financial structures used to fund data centers.

On November 16, Michael Barr, a member of the Federal Reserve’s Board of Governors, addressed these threats during an international conference on artificial intelligence and finance held in Milan.

One of the concerns involves automated systems used to trade stocks and other assets.

Imagine thousands of algorithms analyzing a piece of economic news and simultaneously reaching the exact same conclusion. If everyone decides to sell certain assets, prices could fall rapidly, triggering further automated sell orders.

This behavior can turn an initially limited fluctuation into a much more intense reaction.

According to Barr, machine learning models can also exhibit patterns of coordination or manipulation that are difficult for traditional oversight mechanisms to detect.

The concern is not merely that an artificial intelligence might make mistakes, but that multiple systems could produce similar responses at high speed.

AI agents can create new risks for banks and their customers. Another threat involves so-called artificial intelligence agents—systems capable of performing tasks with relative autonomy.

Unlike traditional chatbots, these programs can execute transactions and interact with digital services on behalf of users.

Barr noted instances where financial institutions identified attempts to access bank accounts without being able to properly verify authorization from the account holders.

Even when a request is legitimate, confirming the identity and permissions of an automated agent poses a challenge.

There is also the risk of coordinated deposit withdrawals.

If numerous agents interpret signs of banking instability in the same way, they could accelerate transfers and withdrawals, increasing pressure on financial institutions.

Other concerns include incorrect recommendations, discriminatory credit decisions, and sophisticated cyberattacks.

These scenarios demonstrate why human oversight and authorization mechanisms remain essential.

Multi-billion dollar investments are also under scrutiny. The expansion of artificial intelligence requires high-cost infrastructure, particularly data centers and specialized equipment.

According to Barr, the financial obligations of the sector’s largest companies are growing rapidly.

Furthermore, complex contracts between technology firms, suppliers, and investors can make it difficult to determine who would bear specific losses.

If growth expectations shift abruptly, these financial interconnections could transmit distress across various market participants. However, the Federal Reserve is not claiming that a bubble is about to burst.

The warning highlights potential vulnerabilities that need to be monitored as the sector grows.

Despite these concerns, the U.S. central bank acknowledges the benefits of the technology.

Artificial intelligence can improve fraud detection, expand access to credit, and increase the efficiency of financial services.

The challenge lies in leveraging these advantages without allowing automated systems to create disproportionate risks.

Key priorities include reliable authorization mechanisms, greater transparency, and oversight of algorithmic decisions.

The Federal Reserve’s message is that innovation must progress hand in hand with appropriate controls.

After all, a technology capable of making financial operations faster can also accelerate the spread of problems when systems do not function as expected.

AI agents can amplify isolated failures or inconsistent data during financial crises by acting autonomously, rapidly, and often in a synchronized manner across interconnected systems.

How minor issues escalate:

• Errors in data and tasks: Incomplete records, unreconciled transactions, or incorrect categorizations can be multiplied by agents that execute financial or fiscal operations without constant human oversight.

• Unpredictable and synchronized behavior: Agents may pursue objectives in ways not foreseen by their creators, or make similar decisions en masse, triggering chain reactions across markets and liquidity.

• Cyber ​​and infrastructure risks: Failures in shared models or attacks carried out by swarms of agents can affect multiple institutions simultaneously—a concern raised by European Central Bank President Christine Lagarde, as noted by GARP.

• Expanded system access: As companies grant more permissions to agents, testing incidents or security breaches can scale up significantly, as pointed out by The Economist.

• Pressure on deposits and intermediaries: Automated searches for higher yields, or a migration toward non-bank finance, can accelerate capital outflows during periods of stress.

 

mundophone

No comments:

Post a Comment

  TECH Researcher finds six previously unknown vulnerabilities in AirDrop and Quick Share A new study from CISPA Helmholtz Center for Infor...