Not yet. As of November 2026, neither the SEC nor the CFTC has adopted a rule written specifically for AI agents. The question is being debated in public: eight members of the US House sent the SEC 13 questions on agentic trading in June 2026, and on 28 October 2026 the CFTC held the first forum in a new public series, on AI and agentic finance.
A note on the legal and regulatory points here: they summarise public sources as of November 2026 and should not be relied on. Rules for AI agents are still being written, and some of what is described here is proposed or coming soon rather than settled. Check the current position, or ask a qualified professional, before acting on it.
“No agent-specific rules” does not mean “no rules”. Existing law still applies to the people and firms involved. What is missing is a rulebook that says how it applies when software makes the decisions.
What has the SEC done on AI so far?
The SEC’s main AI-related rule proposal was withdrawn in 2025, and we found nothing that has replaced it. In July 2023 the SEC proposed rules on conflicts of interest when broker-dealers and investment advisers use predictive data analytics, the kind of technology that includes AI models.
The SEC withdrew that proposal with effect from 17 June 2025. It said it does not intend to issue final rules on it. So as of November 2026, the SEC has no finished rule aimed at AI tools or AI agents.
The pressure is now coming from Congress. On 23 June 2026, eight House members sent the SEC chair a letter with 13 numbered questions about retail platforms letting AI agents trade. They asked about developer liability, broker duties, registration, required safeguards and whether the SEC’s legal authority is sufficient. Answers were requested by 31 July 2026; as of November 2026, we had not located a published response.
What exactly are lawmakers asking the SEC to decide?
The lawmakers are asking the SEC to say how securities law applies at every link in the chain, from the AI developer to the broker to the investor. Their 13 questions fall into four groups, in our summary.
The first group is about what the SEC has already done. Did it consult brokerage platforms or AI firms before agentic trading launched? Did it give any platform or developer formal approval, no-action relief or interpretive guidance?
The second group is about duties. What liability do developers carry when their agents trade for retail investors? Does using a third-party agent reduce a broker’s obligations, such as best-interest, supervision, record-keeping and best-execution duties? When must an agent or its developer register?
The third group is about specific safeguards. The letter lists funding limits, order-size limits, trade approval, transaction logs, dispute processes and limits on an agent’s access to customer data. The fourth is about market-wide risk: herding, manipulation, prompt injection, and whether the SEC needs new powers from Congress.
What has the CFTC done?
The CFTC has gone further than the SEC, but only with consultation, staff guidance and public debate, not new rules. The steps so far, in order:
January 2024 — a question to the market. CFTC staff issued a request for comment on AI in CFTC-regulated markets. It asked how AI is used in trading, risk management and compliance, and about risks such as manipulation, fraud and concentration. Comments closed on 24 April 2024.
December 2024 — a staff advisory. Four CFTC divisions issued a staff advisory on AI use by registered firms. It reminded registrants that existing duties still apply, including when AI comes from an outside provider. It said it creates no new binding rules. The then-chairman called it “a measured first step”.
September 2026 — a forum announced. The CFTC announced a Frontier Forum series, with the first on AI and agentic finance.
October 2026 — the forum held. On 28 October 2026, the agenda included a session on agentic trading. It asked whether and how the CFTC’s framework should apply when agents transact for market participants.
At that forum, the CFTC chair’s remarks, listed on the CFTC’s speeches and testimony page, set out the open questions. Does an agent’s developer have to register? Who is liable if an agent goes rogue? How can regulators know whom an agent is trading for? Is a “know-your-agent” check needed? He stated these views were his own and not necessarily the Commission’s, and he announced no rule or proposal.
So which rules apply to AI agents today?
The rules that apply today are the existing ones, aimed at the people and firms behind the agent. The CFTC staff advisory makes this explicit for registered firms. They must meet their obligations whether they deploy AI themselves or through a third party.
The CFTC chair described the approach as grounded in the functions performed and the control exercised. In plain terms, regulators look at what is actually being done and who is in charge of it, not at whether a human or software pressed the button. That is his stated framing, not a rule.
The chair also pointed to the CFTC’s last attempt at automated-trading rules. Regulation Automated Trading, proposed in 2015, was never finalised. He said it would be ill-suited to today’s landscape. That suggests any future agent rules would start fresh rather than revive the old proposal, though that is our inference.
| Agency | Agent-specific rule? | Most recent formal step | Status as of November 2026 |
|---|---|---|---|
| SEC | No | 2023 AI-related proposal withdrawn, June 2025 | Lawmakers’ 13 questions pending a public answer |
| CFTC | No | Staff advisory on AI, December 2024 | Public forum held 28 October 2026; no proposal announced |
What could go wrong?
The main risk is treating signals as rules. A staff advisory reflects the views of staff divisions and does not bind the Commission. A speech reflects one official’s views. Neither gives you a legal protection you can rely on.
The second risk is assuming that silence means permission. CFTC staff have said existing duties, such as risk management, recordkeeping and protecting customer funds, still apply to registered firms using AI. A product being unregulated as an “agent” does not make it unregulated.
The third risk is speed. Agent tools are launching faster than rules are written. Any product that implies a regulator has approved its agent deserves scepticism. The lawmakers specifically asked whether the SEC had given any platform or AI developer formal approval or relief. A registration, where one exists, is a licence to operate, not an endorsement of the agent’s quality. That last point is our judgement.
The fourth risk is fraud dressed up as innovation. Regulators have not waited for agent rules to act on scams. A CFTC customer advisory on AI trading bot scams warns that fraudsters use AI hype to sell bots promising unreasonably high or guaranteed returns. It describes one scheme in which more than $1.7 billion in bitcoin was taken from at least 23,000 people.
The trust takeaway
There is no agent rulebook yet, so the protections you have are the ones attached to the firms you deal with. Before trusting an agent with money, check:
- Who is regulated. Is the platform, adviser or broker registered with the SEC, the CFTC or the NFA?
- Who runs the agent. Is it the registered firm, an outside provider, or you?
- What the terms say. Who carries the loss if the agent makes a mistake?
- What is official. Is a claim based on a final rule, a proposal, staff guidance or a speech?
- The date. This area changes month to month. Check whether anything has moved since November 2026.