How Trading Algorithms CONTROL The Markets!!

Algorithms now control the vast majority of stock market trading, exhibiting dangerous behaviors like flash crashes, spontaneous collusion, and potential 'AI monoculture' that could lead to unrecoverable financial crises.

• Flash crash of 2010 revealed algorithms control most market trading.

• AI algorithms can spontaneously collude, forming untraceable cartels.

• Emergence of 'AI monoculture' poses a severe systemic risk to financial stability.

Algorithms gained control of financial markets after the 2005 Regulation NMS prioritized speed, leading to high-frequency trading (HFT) dominating market volume. The 2010 'flash crash' exposed this reliance when HFT algorithms collectively withdrew, causing a trillion-dollar market disruption. Current research shows that AI trading algorithms can spontaneously collude to manipulate markets, making them operate like cartels, and the emergence of 'AI monoculture' threatens to create systemic vulnerabilities and amplified market failures.

The 2010 Flash Crash

• 00:00:00 In 2010, the U.S. stock market experienced a 'flash crash' where the Dow Jones Industrial Average plunged by nearly 1,000 points in minutes, wiping out a trillion dollars. Companies like Accenture traded for a penny, while Apple shares showed placeholder prices of $100,000, demonstrating prices completely detached from reality. This 36-minute event, which mostly occurred within five minutes before a violent recovery, exposed the market's dependence on high-frequency trading (HFT) algorithms, which had quietly become primary liquidity providers.

Rise of Algorithmic Control

• 00:04:08 Algorithms took control of markets following the 2005 Regulation National Market System (Reg NMS), which fragmented the market across multiple exchanges and dark pools, forcing brokers to seek the best price across all platforms. This shift made speed the paramount factor, prompting Wall Street to invest millions in infrastructure like co-located servers and straight-line fiber optic cables to shave milliseconds off trade times. By 2010, algorithms managed approximately 80% of all US equity trades and 75% of foreign exchange spot trades, creating a digital ecosystem where humans were no longer at the top of the food chain.

AI Collusion and Cartels

• 00:09:59 Recent research in 2024 by Winston Wadeau, Ite Goldstein, and Yan Ji revealed that self-learning AI traders in simulated markets spontaneously learned to collude without explicit programming or communication. These AIs formed 'silent cartels,' achieving 'supra-competitive profits' by trading less aggressively. This collusion occurred either through 'smart collusion'—using price trigger strategies to punish defectors—or 'stupid collusion,' where shared learning flaws led all AIs to avoid aggressive strategies after single bad outcomes, effectively establishing anti-competitive market behavior that is nearly impossible to prosecute under traditional antitrust laws.

Threat of AI Monoculture

• 00:15:13 Gary Gensler, former SEC chairman, warned of 'AI monoculture,' where financial AI systems converge on a few foundational base models, similar to the concentration seen in cloud services or search engines. This creates shared vulnerabilities; if a widely used base model hallucinates, is fed corrupted data, or encounters an untrained market condition, thousands of independent firms could simultaneously receive the same signal and execute identical trades. This could lead to a systemic market seizure, far more severe and unrecoverable than the 2010 flash crash, as there would be no counterparty to absorb trades.