AI and machine learning in automated trading

The automated trading landscape is evolving from static, rule-based scripts toward autonomous AI-driven agents. These next-generation systems utilize machine learning to analyze complex, multi-source data—including liquidity shifts and on-chain movements—to function more like human strategists than simple calculators.

As these technologies advance, there is an increasing focus on using artificial intelligence and machine learning to manage execution costs. Retail traders are employing these models to identify and mitigate hidden costs such as spreads, slippage, and execution delays. In specialized environments like Polymarket, managing slippage through order-book depth estimation and order splitting has become a critical component of maintaining profitability.

This shift toward agentic trading has introduced significant regulatory and liability concerns. As brokers connect AI tools to client accounts, uncertainty remains regarding responsibility for costly autonomous errors. Robinhood has maintained that “customers remain responsible for the actions of their agents,” even as over 50,000 of its customers have opened agentic trading accounts to trade millions of dollars daily in equities and options.

To address these risks, experts are discussing potential safeguards such as kill switches, circuit breakers, and ‘Know Your Agent’ rules. While legal and regulatory experts anticipate that firms will eventually face greater obligations as the technology matures, the current landscape remains focused on defining the boundaries of accountability between users and autonomous systems.