AI agents have been connecting to brokerage platforms for barely six months, yet regulatory frameworks remain absent as authorities grapple with how to oversee this emerging technology. Bank of England Deputy Governor Sarah Breeden acknowledged in June that traditional human-in-the-loop safeguards are impractical given the speed of agentic trading, suggesting instead that regulators may look toward kill switches and circuit breakers borrowed from algorithmic trading oversight.
Several brokers have already moved forward despite regulatory uncertainty. ThinkMarkets launched Chelsea AI in June, a plugin allowing AI models like Claude and ChatGPT to control trading accounts directly. Robinhood, IG Group and eToro have followed suit, while infrastructure providers MetaQuotes and Spotware have released their own versions, making deployment straightforward across the industry.
In the European Union, ESMA issued supervisory guidance in February covering AI-based algorithmic trading systems, though the EU AI Act itself lacks specific provisions for agentic trading. The UK’s Financial Conduct Authority published the Mills Review in July discussing how AI will transform financial services by 2030, but stopped short of addressing agentic trading directly.
The technology remains immature. Recent public trading contests featuring leading AI models from OpenAI and Alibaba showed most lost money and executed excessive trades, raising fundamental questions about liability when agents malfunction. ThinkMarkets CEO Nauman Anees notes that implementing and enforcing AI governance rules would be extremely difficult at this stage.
FXnCO Insight
Brokers deploying AI trading agents should implement robust risk controls and maintain adequate capital buffers now, as regulatory clarity will likely follow only after market failures expose gaps.
Source: Finance Magnates