Artificial intelligence is having its loudest month yet. The executives building the most advanced systems are publicly calling for stronger oversight, lawmakers on several continents are writing binding rules, and polls show most Americans are uneasy about where this is heading.
This article is our contribution to that conversation: an honest look at what people are worried about, what the rules now require, and how we designed AI Portfolios around a simple principle. AI should do the analysis, while a human makes every decision that moves money.
The month the debate sharpened.
In the second week of September 2026, the CEO of Anthropic published an essay titled “We Must Pace the Frontier,” arguing that AI capabilities are accelerating faster than our ability to oversee them. Days later, the CEO of OpenAI agreed, and delayed his company’s IPO to 2027 on safety grounds. A 27-year-old AI researcher resigned with a public warning that was viewed 26 million times. Then both CEOs stood before the UN Security Council and warned that poorly managed AI “could threaten humanity.”
Whatever you make of the strongest warnings, the headlines marked a shift: the people building these systems are now publicly discussing the need for stronger oversight.
The public felt it first. In June, Pew found that 71 percent of American adults expect AI to reduce the number of jobs over the next 20 years, up from 64 percent two years earlier. In September, Data for Progress found 68 percent of voters support pausing the development of advanced AI, with Democrats, independents, and Republicans in unusual agreement. A CNN poll the same month found a majority of Americans feel “fear and concern” about AI.
It’s not the intelligence, it’s the accountability.
Listen closely to what people are actually saying, and a pattern emerges. The fear is not that machines are smart. It’s that no one is in charge of them.
For workers, it’s the fear that a system will decide their job is redundant and no human will second-guess it. For parents, the concern is urgent: lawsuits filed this year alleged that chatbots discussed suicide methods with teenagers, and one family described a chatbot as the only confidant their child felt understood him. California responded in September with Adam’s Law, the toughest chatbot child-safety rules in the country.
The rules are learning to ask the same question.
Here’s what changed in 2026: the rules stopped being proposals. The EU’s AI Act switched on its enforcement machinery in August, with fines reaching up to 3 percent of a company’s global turnover. With no comprehensive federal law in sight, American states started writing binding statutes of their own: Texas, California, New York, Connecticut.
And in finance, every regulator that matters is converging on one demand. The SEC’s 2026 examination priorities name AI fiduciary duty outright, and examiners will check whether firms’ claims about their AI are accurate. FINRA holds that its rules are technology-neutral, which is a polite way of saying the firm is fully responsible no matter what the software did, and its 2026 priorities specifically call out AI agents trading stocks. The Financial Stability Board told banks to treat capable AI agents like “synthetic employees,” with human approval above thresholds and kill switches. A Senate draft, the AI AGENT Act, proposes non-waivable duties of care and loyalty for AI agents. Europe requires human oversight built into the architecture of high-risk systems. South Korea requires meaningful human oversight “at any time” for AI in finance.
Different continents, different legal traditions, same question: who is accountable?
Now make it about your money.
This is where the abstract debate gets personal. An AI can be wrong about a trivia question and it’s funny. Wrong about your retirement and it’s your life.
People already sense the difference. In July, Bentley and Gallup found the public rates AI as no better than people at recommending financial advice, and nearly a third said they don’t trust businesses at all to use AI responsibly. That’s not a public waiting to hand its savings to an algorithm. It’s a public waiting for someone to be accountable.
So here’s our honest answer to the question underneath this whole debate, the one a skeptical reader is really asking: why should I trust an AI with my money?
You shouldn’t trust an AI with your money. You should trust a system where AI does the analysis and a human makes every decision that moves a dollar.
How we built for this.
Our founder(s) started AI Portfolios after seeing an increasing demand for always-on financial advice and the desire for simplified investing. From the beginning, the design principle was the one regulators are now writing into law: AI Portfolios is built so AI does the heavy lifting on research and proposals, while a licensed human stays accountable for every action that moves money.
Here’s what that looks like in practice.
Director, our always-on advisor, reads market conditions every day and translates what it sees into plain English, with a published confidence score and an explicit “not a forecast” stance. When it hasn’t measured something, it says so. Sage, its conversational side, lets you review your portfolio and ask questions any time, even at midnight.
When it’s time to act, Autopilot does one thing: it proposes. They rank what to buy, sell, or rebalance. That is the full extent of their power. You review each recommendation and approve the trades you want to make. Nothing moves without your say-so. That’s the first layer of security.
The second layer happens on the Firm Desk, our control surface. An registered financial advisor reviews every proposal you approved and clicks Approve or Deny. Nothing bound for the custodian leaves our system without that human decision, and where your consent is required, it applies too. If either layer says no, nothing moves.
Two layers of security on every trade, every time.
That gate is permanent. It isn’t a cautious pilot setting we’ll relax once everyone feels comfortable. Straight-through processing and automatic execution are explicit non-goals. AI can be aggressive on analysis and proposals; it still can’t touch a dollar without a human decision, enforced in software, not just a policy document.
We never hold your money. Your account at IBKR or Schwab stays your account, with the same custodian you already trust. We never see or store your brokerage passwords. You link your account through your institution’s own secure sign-in.
You set the guardrails: target allocation, exclusions, maximum position size, tax sensitivity. We harvest tax losses where it helps. You can stop your activity at any time; separately, our firm-wide kill switch withholds exports even for items approved earlier. Portfolios are built the boring, responsible way: low-cost, diversified ETFs at the core, individual stocks for growth, matched to your risk profile.
Behind the scenes, the same principle runs through the whole platform. Signing in is tied to a verified identity, sensitive details stay locked away, and the practice environment can never touch real money. None of this is the interesting part of the story, which is exactly the point. Safety should be uneventful.
What responsible AI in finance looks like.
If we had to write the standard the regulators are circling, it would be five lines long:
- AI proposes and humans dispose. Every action that moves money has a named human accountable for it.
- No custody, no exceptions. Your money stays with your custodian.
- Explain yourself. Plain English, published confidence, and the humility to say “not a forecast.”
- Controls the client actually holds: guardrails, pause, stop, and a kill switch.
- The gate lives in the software, not just the policy doc.
That’s the whole philosophy. Human-in-the-loop isn’t a UI flourish or a marketing slogan. It’s the control.
Frequently asked questions about AI wealth advisors.
Does the AI ever trade on its own?+
No. Our AI systems only propose trades. You approve the ones you want, then an advisor reviews and approves them on the Firm Desk before anything is sent to the custodian. Automatic trading is not something we offer, now or later.
Who is accountable for my trades?+
A licensed human is accountable for every action that moves money. The AI does the research and drafting; a person checks and approves.
Do you hold my money?+
No. Your account stays at your custodian, IBKR or Schwab. We never hold client funds. We also don’t hold your brokerage passwords; you connect your account through your institution’s own secure sign-in, and we never see or store your passwords.
What if I want to stop trading?+
You can stop your activity at any time. Separately, we have a firm-wide kill switch that halts exports, even for items that were already approved.
What if the AI gets something wrong?+
Director shows a confidence score with its market reads, says plainly when something is not a forecast, and says so when it hasn’t measured something. And because no trade happens without your approval and an advisor’s approval, a wrong read can’t become a wrong trade on its own.
Can I set limits on what the AI recommends?+
Yes. You set guardrails like target allocation, exclusions, maximum position size, and tax sensitivity. We also harvest tax losses where it helps.
Experience AI Portfolios for yourself.
The debate this month isn’t a reason to fear the technology. It’s a reason to expect accountability, and to look for companies that treat accountability as the product.
We built the thing we wished existed: the analytical power of AI, with a human hand on the wheel and your money exactly where you left it.
Come see what accountable AI feels like. Join the early access list for AI Portfolios today.