Coinbase expands AI trading to stocks
Coinbase (NASDAQ: COIN) is rolling out AI agent trading for securities, CEO Brian Armstrong announced on X earlier today. The service was previously limited to crypto and derivatives. The move puts one of the largest U.S. crypto exchanges directly into the emerging market for automated stock trading - a shift Armstrong describes as the beginning of "AiFi," or agentic finance.
Supported AI agents include ChatGPT, Claude, and Cursor, with Codex coming soon, according to Coinbase's documentation. Supported equities cover S&P 500 U.S. stocks. The company has been signaling this expansion for months, having previously disclosed its intent to enable AI agent trading.
Coinbase is not treating AI as a fiduciary. The company's documentation states that AI agents can make errors and does not guarantee the accuracy of any action taken by an AI agent. "You are solely responsible for reviewing and authorizing any trades, transfers, or account changes made through agentic workflows."
Who else is in the race
Coinbase has company. Robinhood, Public, eToro, Interactive Brokers, and Gemini/Bitget are all offering similar services or moving in the same direction. The competitive field is forming quickly, but the products are still new and usage data remains thin.
Some sources have claimed around 500,000 active agents on Coinbase. Robinhood topped 50,000 agentic accounts in the first weeks after its May 27 launch and reportedly exceeded 100,000 by its late-July earnings report. These numbers are early and unaudited, but they suggest real demand for letting software execute trades.
For finance professionals tracking AI Agents & Automation, the pattern is familiar: retail platforms ship the feature first, usage follows, and institutional tooling arrives later. The difference here is that the "retail" users are increasingly technical - developers wiring ChatGPT or Claude into brokerage APIs.
Where the risk sits
Coinbase's liability language is blunt. The user reviews and authorizes every action. That shifts the burden of mistakes - bad trades, wrong quantities, unintended transfers - entirely onto the account holder. For anyone in a compliance or risk role, this is the central tension of agentic finance: the agent acts, but the human signs.
No major regulator has issued specific guidance on AI-executed securities trades. The SEC's existing rules on algorithmic trading and best execution were written for institutional desks, not consumer-facing language models connected to brokerage accounts. That gap will likely close, but not before some agents make costly errors in public view.
Why this matters for finance professionals
If you work in wealth management, trading, or fintech operations, agentic finance is no longer a concept deck item. Coinbase and Robinhood now have live products. The practical question is whether your firm's risk controls can distinguish between a human clicking "buy" and an AI agent doing it through an API. If the answer is no, that's the first gap to close.
Your membership also unlocks:
