News · Analysis
AI Trading Agents Outpace Regulation on Liability
Six months after brokers began connecting AI agents to trading platforms, no regulatory framework yet defines who is liable when an agent misfires, according to Finance Magnates (17 August 2026). Bank of England Deputy Governor Sarah Breeden said in June that human-in-the-loop safeguards are unrealistic for agentic trading speeds, pointing instead to kill switches and circuit breakers, per the report. Robinhood has told Finance Magnates that customers are solely responsible for how agents are built and the actions they take, though the report notes lawyers doubt regulation will place all responsibility on traders. For now, brokers and traders are operating in a gap between rapid product rollout and slow-moving policy.
Brokers have spent the past six months wiring large-language-model agents directly into trading accounts, letting tools such as Claude, Gemini and ChatGPT place trades on a client's behalf. According to Finance Magnates (17 August 2026), the pace of adoption has outstripped any settled view among regulators, lawyers or brokers on a basic question: who pays when the agent goes wrong.
What happened
Finance Magnates reports that ThinkMarkets was an early mover, releasing its Chelsea AI plug-in in June 2026, before Robinhood, IG Group, eToro and infrastructure providers MetaQuotes and Spotware introduced their own versions. Robinhood told the outlet that "customers are solely responsible for how the agents are built and any resulting actions the agents take." Meanwhile, Bank of England Deputy Governor Sarah Breeden said during a Q&A at the European Central Bank Forum in Portugal in June 2026 that human-in-the-loop review — where a person can check, approve or amend an automated action — is unrealistic given the speed of agentic trading. Instead, per the report, regulators are examining kill switches and circuit breakers, tools already familiar from algorithmic trading oversight. In the EU, ESMA's February 2026 supervisory briefing on "AI-based algorithmic trading systems" indicated such systems would fall under the EU AI Act, though Finance Magnates notes the Act has no distinct category for autonomous agents. The UK's Financial Conduct Authority published its Mills Review in July 2026, addressing how AI could reshape financial services by 2030, but the report states it did not directly address agentic trading.
Why it matters
ThinkMarkets chief executive Nauman Anees told Finance Magnates that "there is no AI governance or rules and any would be difficult, to say the least, to implement and enforce." That absence of a framework leaves an open question over accountability: if a trader's agent executes an unauthorised or erroneous trade, current statements from firms such as Robinhood place the burden on the customer, yet the report suggests future "Know Your Agent" rules could shift some responsibility back to firms — adding compliance costs in the process. Until regulators settle this, both brokers and traders are exposed to ambiguity that ordinary terms of service may not fully resolve.
What it means for traders
Traders using agentic plug-ins should treat provider disclaimers carefully, given Robinhood's stated position that account holders bear responsibility for an agent's actions. This is a useful moment to consider a broker's wider regulatory standing and governance track record before delegating trade execution to an automated system; readers weighing this can see our Best Regulated Broker winners for context on how regulatory credentials are assessed, though inclusion there does not imply any endorsement of the firms named in this report. Our approach to evaluating brokers more generally is set out in our methodology.
What to watch
Three threads are worth monitoring, per Finance Magnates: further comment from the Bank of England on kill switches and circuit breakers as agentic-trading safeguards; whether ESMA clarifies how the EU AI Act applies to autonomous trading agents specifically; and any follow-through from the FCA's Mills Review into rules that address agentic trading directly, alongside prospective "Know Your Agent" obligations for brokers.
FAQ
Who is liable if an AI trading agent misfires? According to Finance Magnates (17 August 2026), there is no settled liability framework yet. Robinhood has stated that customers are solely responsible for how agents are built and the actions they take, though the report notes that lawyers doubt regulation will ultimately place all responsibility on traders.
What safeguards are regulators considering instead of human review? Bank of England Deputy Governor Sarah Breeden said at the ECB Forum in Portugal in June 2026 that human-in-the-loop review is unrealistic for the speed of agentic trading, and that kill switches and circuit breakers are being examined instead, per Finance Magnates.
Does the EU AI Act currently cover agentic trading? ESMA's February 2026 supervisory briefing indicated AI-based algorithmic trading systems would need to comply with the EU AI Act, but Finance Magnates reports the Act does not yet have a distinct category for AI agents.