US Lawmakers Press SEC on Guardrails for AI ‘Agentic Trading’

Members of the US Congress have raised fresh concerns about AI‑driven “agentic trading” — autonomous trading carried out by AI agents on behalf of retail investors. In a 23 June 2026 letter to the Securities and Exchange Commission (SEC), lawmakers warn that brokerage platforms are beginning to let external AI agents plug directly into customer accounts, and ask whether existing securities rules are sufficient to address the new risks posed by generative‑AI‑based trading tools.

The letter requests detailed written responses from the SEC by 31 July 2026. Among the questions: Has the SEC granted any formal approvals or no‑action relief for platforms offering agentic trading? Should there be hard limits on funding, position size or order size for AI‑driven strategies? What obligations should apply around transaction logging, dispute resolution and restrictions on model access to sensitive customer data? Crucially, lawmakers ask whether a broker‑dealer’s duties — such as best‑execution, supervision, and cybersecurity — remain fully in force when a client chooses to use a third‑party AI agent to place trades.

Algorithmic and high‑frequency trading have long been part of US markets, but the spread of large‑language‑model‑based agents to retail users raises novel governance issues. Recent academic work has highlighted how current rulebooks, including the EU AI Act, were largely written with more static AI systems in mind and may struggle to capture the behavior of adaptive, multi‑step agents that can initiate transactions or modify strategies on their own. For regulators in Japan and elsewhere, the US debate foreshadows a coming wave of questions about how to supervise AI that is not just recommending trades, but actually acting in investors’ names.