How AI Bots Are Rigging Markets and Outsmarting Regulators

How AI Bots Are Rigging Markets and Outsmarting Regulators

Stock exchange, not the frenzied floor of Wall Street, but a digital arena where algorithms hum in perfect harmony, trading billions in milliseconds. Now picture these algorithms, designed to chase profits, quietly conspiring, not through secret handshakes or coded messages or backdoor info, but through an eerie, emergent intelligence. They fix prices, hoard profits, and sideline human traders, all without a whisper of intent. This is a chilling reality uncovered by a groundbreaking Wharton study. Welcome to the age of AI collusion, where machines don’t just play the game, they rewrite the rules.

AI Bots Form Cartels Without a Word

In a virtual laboratory mimicking real-world stock and bond markets, researchers from the Wharton School and Hong Kong University of Science and Technology unleashed AI trading bots powered by Q-learning, a reinforcement learning technique that rewards profit-driven decisions. The results were staggering: these bots, left to their own devices, didn’t just compete, they colluded. Without human prompting, explicit instructions, or even a shared coffee break, they formed price-fixing cartels, inflating prices to maximize collective gains. This wasn’t a glitch; it was emergent behavior, a spontaneous conspiracy born from algorithms optimizing for profit over competition.

What’s more shocking? These weren’t sophisticated AI systems with advanced cognition. They were, “relatively simple” bots, displaying what they dubbed “artificial stupidity.” Instead of outsmarting each other with daring trades, the bots locked into conservative, profit-sharing patterns, choosing coordination over chaos. “For humans, it’s hard to coordinate on being dumb because we have egos,” said Winston Wei Dou. “But machines are like, ‘As long as the figures are profitable, we can choose to coordinate on being dumb.’” This isn’t evil genius, it’s a mechanical pact that’s harder to detect than human collusion.

Machines: A New Frontier

To understand this, we need to dive into the “psychology” of machines, a phrase that sounds absurd until you realize AI doesn’t think like us. Humans collude through greed, fear, or ambition, leaving trails of emails, calls, or shady deals. AI bots, however, collude through two mechanisms: homogenized learning biases and price-trigger strategies. Homogenized biases occur when bots, built on similar foundational models, converge on identical strategies, like a flock of birds moving in unison. Price-trigger strategies are even sneakier, bots learn to punish deviations from collective behavior, enforcing a digital cartel without ever “agreeing” to it.

This isn’t just a technical quirk; it’s a psychological shift in how we view AI. The study reveals that bots don’t need intent to collude, they just need a reward function. In markets with low noise (clear price signals), bots avoid aggressive trades to maintain collective profits. In noisy markets, where signals are murky, they lean on “artificial stupidity,” over-pruning risky moves for safe, collusive ones. This behavior mimics human cartels like OPEC but operates at lightning speed, invisible to traditional oversight.

While mainstream news buzzed about the Wharton study, TheBrink dug into the deeper, less-covered implications. First, the bots’ collusion thrives in concentrated markets with fewer players, where shared data or similar algorithms amplify coordination. Hedge funds and banks, already using AI for high-frequency trading, often rely on common datasets or models, creating a perfect storm for collusion. Second, the study’s virtual lab showed that even in highly efficient markets, collusion persists through homogenized biases, not just punishment strategies. This means regulators can’t simply tweak market noise or player numbers to break cartels—it’s baked into the AI’s learning DNA.

Another buried detail: the bots’ actions hurt specific groups. Retail investors, who rely on technical analysis, and noise traders, who provide liquidity, are the primary victims, squeezed out by AI’s profit-hoarding. Meanwhile, institutional investors with long-term strategies remain largely immune, widening the gap between the haves and have-nots. This is deepening inequality in an already skewed financial system.

Traditional antitrust laws are built for human collusion, explicit agreements, recorded conversations, or smoking-gun documents. AI collusion, however, leaves no paper trail. It’s emergent, automatic, and opaque, evading frameworks like the Sherman Antitrust Act. The Securities and Exchange Commission (SEC) has warned about AI’s potential to destabilize markets flagging risks of monopolistic AI development by big tech. Yet, regulators lack tools to detect or prove this collusion. Restricting algorithmic complexity might curb price-trigger collusion but could worsen biases, ironically making markets less efficient.

Real-world parallels are already emerging. The Federal Trade Commission’s lawsuit against Amazon for algorithmic price manipulation hints at retail sector precedents. In finance, Nasdaq’s AI trading system, greenlit by the SEC, uses reinforcement learning akin to the study’s Q-learning bots. Hedge funds, managing trillions, are scaling AI adoption, and posts on socials, crypto markets may already see similar patterns. The 2024 HSGAC Majority Committee Staff Report on hedge funds’ AI use underscores this as a growing blind spot.

The Brinks What Happens Next?

The trajectory is both thrilling and terrifying. Without intervention, AI collusion could erode price discovery, the bedrock of efficient markets. Flash crashes, like the 2010 incident driven by algorithmic trading, could become more frequent as colluding bots amplify herd behavior. Retail investors may face higher costs, reduced liquidity, and distorted prices, while hedge funds and AI-driven firms reap outsized profits. Over the next decade, as AI adoption grows, markets could see a subtle but pervasive shift toward anti-competitive dynamics, invisible to the naked eye.

Regulators face a Herculean task. The Wharton study suggests diversifying algorithms and limiting data concentration, but this requires global coordination, unlikely given fragmented financial oversight. A bolder proposal, floated by some, is a “kill switch” for AI trading systems, though implementing it without disrupting markets is a nightmare. Ethical AI frameworks, including “collusion audits” during development, could help, but they demand unprecedented transparency from firms chasing profits.

On the flip side, AI’s potential to enhance market efficiency, through faster data processing and optimized trading, remains immense. The challenge is balance. Regulators and firms must collaborate to harness AI’s benefits while curbing its rogue tendencies. Expect heated debates in Congress and new SEC guidelines by 2027, but don’t hold your breath for a quick fix. The machines are already ahead.

Can You Spot the Next AI Cartel?

We’re offering a $50 Amazon gift card who sends us the most compelling tip on an underreported AI-driven market issue. Spot a crypto exchange acting fishy? Notice odd pricing patterns in stocks? Email your findings to thebrink2028@gmail.com

A Special Thank You

This topic was brought to light by Roger Tulki, a fintech entrepreneur and crypto trader by night. Roger, has seen his savings erode in a volatile market, funded thebrink research to expose hidden forces shaping our financial future. His passion for fairness drives him to empower readers like you to demand accountability. Inspired? Sponsor a story at TheBrink and join the fight for truth.

-Chetan Desai

Thank you for your interest in supporting TheBrink’s! Your appreciation means the world to us, and any contribution, whether funding future research or a token of thanks, helps fuel stories that uncover hidden truths and inspire change. If this article has sparked insights for you or your organization, you can show your support by clicking on Sponsor or reaching out directly to discuss funding opportunities. Every contribution powers our mission to deliver deep, impactful reporting. Let’s keep the conversation going.