AI Trading Agents Face Regulatory Scrutiny

The rapid integration of AI agents into retail trading platforms is outpacing regulatory frameworks, raising concerns about accountability and investor protection. Within six months of brokers connecting AI agents to their platforms, regulators are still grappling with how to supervise this new technology.

During a Q&A at the European Central Bank Forum in June, Bank of England Deputy Governor Sarah Breeden highlighted the impracticality of traditional human-in-the-loop safeguards for the speed of agentic trading. Instead, regulators are exploring proven measures from algorithmic trading, such as kill switches and circuit breakers. However, the question of liability remains unresolved: if an AI agent makes a rogue trade, who is responsible?

The regulatory landscape is fragmented. In the EU, ESMA's February supervisory briefing addressed 'AI-based algorithmic trading systems,' suggesting they may fall under the AI Act's requirements, though the Act does not have a specific category for agentic trading. The UK's FCA Mills Review, published in July, discussed AI's broader impact on financial services but lacked direct guidance on agentic trading.

Meanwhile, brokers are moving ahead. ThinkMarkets released Chelsea AI in June, enabling traders to let Claude, Gemini, ChatGPT, and other models manage their accounts. MetaQuotes and Spotware, which dominate the industry's infrastructure, have also introduced their own versions. Robinhood launched AI Agent Accounts in July, attracting over 50,000 customers within a month, with daily trading volumes in equities and options.

Early real-money contests with AI trading have shown mixed results, with models incurring losses and making erratic decisions. This unpredictability underscores the need for oversight. As the industry evolves, retail traders must understand that AI agents operate in a largely unregulated space, and protections are still in development.

This is information, not investment advice.