Robinhood's AI Trading Agents Are Here. Nobody Can Tell You Who's Liable.
Robinhood Just Gave 29 Million People an AI Trading Agent. Who Pays When It Loses Your Money?
Let me ask you a question, and I want you to answer it honestly. Would you hand a chatbot your savings account, tell it to "look for opportunities", and then go to bed?
Because as of this week, roughly 29 million Robinhood customers can do exactly that. And the part that bothers me most isn't the technology. It's that when I went looking for the answer to "who pays if it goes wrong", the honest answer was; nobody knows.
What Robinhood Actually Launched
Here's what Fortune reported. Robinhood is rolling out trading agents, powered by OpenAI and Anthropic models, to its entire customer base. You give your agent a name, pick a model (OpenAI's GPT-6 Luna, GPT-6 Sol, or Anthropic's Opus 4.8), and then talk to it in plain English. Simple stuff, like "Buy $200 of Ford stock". Or something bigger, which Robinhood calls a "Loop"; a standing instruction that checks the market every morning and trades when conditions are met, or runs a continuous overnight strategy to look for opportunities while you sleep.
Robinhood built in some guardrails. The agent gets its own dedicated trading account, you can set limits on how much it trades at a time, and you can opt in to a confirmation step where the agent has to get your approval before it executes. Robinhood's CEO says the goal is to give everyday people tools once reserved for hedge funds, big banks, and quant firms.
And this isn't a test run. Fortune reports more than 150,000 customers already opened agentic accounts through the more technical version this spring, and agents are now making nearly 30 million transactions a day on the platform.
Read That Guardrail List Again
I'm going to be direct about this, because I think the friendly language hides something. The confirmation step, the one that makes the agent ask before it acts, is optional. The limits are set by the user, the same user who is being told this tool is as good as a hedge fund's. And the flagship feature is the one designed to run when you're not looking.
A hedge fund has risk desks, compliance officers, and a team of people whose entire job is to pull the plug. You have you, a phone, and a Loop running at 3 a.m. Calling that the same toolkit is a stretch. I don't think it's a lie. I think it's a sales pitch, and you deserve to hear the difference.
The Question Nobody Has Answered: Who Is Liable?
This is the heart of it for me. Fortune says it plainly; if something goes wrong with agent-based trading, it's unclear where legal liability would fall. Robinhood's position is that hosting these agents isn't financial advice, and that what the agent suggests or does is like asking the internet or a friend. The article also notes the legal landscape is still evolving.
Think about what that framing means. If a friend tells you to buy a stock, and it tanks, nobody sues the friend. But this friend has access to your brokerage account, can place trades on its own, and is sold to you by the platform holding your money. That is not the same as a buddy at a barbecue.
Now, I'm not a lawyer, and I want to be precise here because precision over panic is how I do this. Regulators do say their existing rules apply to AI. The SEC and FINRA have stated that their obligations around supervision, fraud, and fiduciary duty don't get a grace period just because the tool is new. But those rules mostly land on registered brokers and advisers who use AI to serve clients. Whether a platform that says it's merely "hosting" your agent falls under them is exactly the kind of question that hasn't been settled. An SEC commissioner said in a March speech that liability worries are the biggest thing holding back AI adoption among advisers, and that there are real regulatory questions about whether it counts as marketing, whether it needs registration, and how it would be supervised. If the professionals are asking those questions out loud, you should be too.
And Here's the Part That Hits Your Wallet
A lot of people assume there's a safety net. There is one, but it's not the one you think. SIPC, the insurance that comes with a U.S. brokerage account, protects your cash and securities if the brokerage itself fails. It does not protect you against losses from a decline in the market value of your securities, and it doesn't cover you against bad investment advice.
So put the pieces together. The agent makes a trade you didn't expect. Or it follows your instruction to the letter in a market that turned. The money is gone. The platform says it wasn't advice. The insurance says it doesn't cover market losses. And the AI company says it only supplied the model. That's the loop I'd worry about, and no setting inside the app closes it.
You lose your money. Everyone else in that chain has an explanation.
The Ethical Questions I Can't Shake
If an agent makes a decision no human reviewed, and it costs someone their retirement, whose decision was it?
Fortune raised the scenario of many agents, trained on similar data, moving into or out of the same asset at once. Who's responsible for a herd of machines, and who's watching it?
Robinhood's exec team says the cost of running these agents will be negligible for most people. What happens if compute costs rise, or if people run research-heavy strategies, as the article itself notes?
Who benefits when millions of people trade more often? Ask that one every time a company tells you it's democratizing something.
I'm not claiming bad intent. I'm saying these are the questions that get asked before you hand over the keys, not after.
Look at the Pattern
If you've been with me for a while, this should feel familiar. On the WISeR Medicare program, an algorithm with a financial incentive was deciding who got care, and the accountability was foggy. In the AI safety warnings I've covered, the people building this technology said they didn't have a full plan for controlling it. Different industry, same shape; powerful automation goes out first, and the question of who answers for it gets sorted out later, usually by the people who got hurt.
If You're Tempted, Here's How I'd Think About It
I'm not a financial advisor, and this isn't advice. This is just what I'd want to know before touching it.
Never put in money you can't afford to lose. Treat whatever you fund the agent with as gone.
If you try it at all, turn the confirmation step on so the agent has to ask before it acts. Don't run Loops overnight.
Set the tightest trading limit the app allows, and lower it, not higher, when you feel confident.
Read the terms and disclosures before you click agree. Look for anything that says who is responsible for the agent's actions. If you can't find it, that tells you something.
Keep your retirement accounts out of it. Long-term savings and an experimental overnight bot shouldn't share a room.
And before you decide, ask yourself the same thing I asked at the start; if this goes wrong, and it might, who is going to make me whole? If the honest answer is "no one", you already know what to do with that.
Where I Land
I'm not saying AI has no place in finance. Tools that help you understand your own money, flag a suspicious charge, or explain what a fee is have real value. But there's a canyon between a tool that informs you and a system that acts for you with your money, while you're asleep, with the liability question left blank.
Question this. Ask for the answers in writing. Make the companies say out loud who is responsible, because right now they don't have to.
Stay sharp. Stay loud.
Michael Routhier is the founder of Tech 4 Grown-Ups, providing honest, unfiltered digital literacy for adults 55+, and host of The Virtuous Machine, exploring the ethics and human cost of AI. Read by tech-curious readers in 200+ countries. Explore more at tech4grownups.com.
Comments