Three Brokerages Gave AI the Keys. Three Others Locked the Door.

Six major brokerage and trading platforms have launched AI-powered trading tools within weeks of each other in June 2026, according to a Jefferies industry summary reported by Investing.com.

What makes the moment worth noting is not just the launches. It is how differently each company answered the same underlying question: how much authority should an AI agent actually have over a customer's money.

The six platforms split into three distinct tiers of trust, and the gap between them is wide enough to function as a real-time experiment in financial AI risk tolerance.

Full Autonomy

At the most permissive end of the spectrum, Coinbase, eToro, and Robinhood have all introduced fully automated AI agents that can make trade recommendations, build portfolios, and execute trades automatically, constrained only by limits the user sets in advance. 

All three platforms are built on Model Context Protocol servers, the open standard that allows large language models such as Anthropic's Claude or OpenAI's GPT to connect directly with a company's internal trading infrastructure through a standardized interface.

Robinhood's approach requires customers to open a dedicated agentic trading account, separate from a standard brokerage account, and currently supports only equities, with additional asset classes planned for the future. 

The company disclosed on June 18 that more than 50,000 customers opened agentic trading accounts within the first few weeks of availability, with those accounts trading millions of dollars daily across equities and options. 

That adoption curve, in a product category that did not exist a month earlier, is among the fastest documented rollouts of autonomous trading access at retail scale, according to Jefferies.

eToro's agent, branded Tori, supports a notably broader range of asset classes than Robinhood's offering including equities, commodities, cryptocurrency, exchange-traded funds, and foreign exchange. 

The company reported on May 12 that Tori has facilitated more than 500,000 trades and is now used by more than one-third of eToro's club members within its first year of availability, suggesting deeper penetration into an existing user base rather than purely new account acquisition.

Coinbase's version is scoped specifically around crypto, offering support for spot trading and derivatives with plans to expand into additional asset classes. 

At the company's System Update event on June 16, Coinbase disclosed that its agent platforms generated more than $4 million in revenue through 40,000 agents operating on Virtuals, and more than $30 million in agent earnings on Banker.

That same week, Coinbase took the autonomy question further. The company launched Coinbase Advisor, registering it simultaneously with the SEC, CFTC, and NFA, the first AI agent of its kind to hold all three credentials at once. 

Under the Investment Advisers Act of 1940, that registration makes Coinbase Advisor a fiduciary, legally required to put the customer's interests first in every recommendation. 

But Coinbase's own disclosures state that the adviser's output "may be inaccurate or incomplete" and that investment outcomes remain the customer's responsibility. 

A platform claiming fiduciary duty while disclaiming responsibility for the outcomes of that duty is a combination with no precedent in securities law, and it has not yet been tested in court.

Supervised Autonomy

Interactive Brokers occupies a narrower middle position. The firm's semi-automated tool, launched in June, can research stocks and surface real-time market insights with the same analytical depth as its fully autonomous competitors. 

But every trade requires manual approval before execution, under what the company describes explicitly as a "human in the middle" strategy. The offering currently supports equities and ETFs.

The firm is not withholding AI capability from its platform. It is withholding execution authority specifically, treating trade approval as a function that should remain human regardless of how capable the underlying recommendation engine becomes.

No Trading Authority At All

At the most conservative end of the spectrum, Charles Schwab and Tradeweb are building conversational AI agents that can query market data and generate portfolio summaries, but are explicitly prevented from making trade recommendations or executing transactions of any kind. 

These tools function closer to an intelligent research assistant than a trading agent in any meaningful sense.

Schwab's version is not expected to launch until the second half of 2026, and will support a notably broad range of products once it does, including equities, options, fixed income, mutual funds, and ETFs. 

Tradeweb’s offering currently supports US credit markets, with plans to expand into global credit and government bonds.

The contrast with Robinhood, eToro, and Coinbase is stark. Schwab and Tradeweb are operating in adjacent product categories that carry materially different regulatory exposure and customer risk profiles than equities, options, or crypto trading at Robinhood, eToro, or Coinbase's retail scale. 

Their caution is not necessarily a statement about AI capability. It may be a statement about what those specific markets and client bases can tolerate.

What This Actually Means

The three-tier split across six companies is not simply six different product roadmaps arriving at different speeds. 

It is six different institutional judgments about the same underlying tradeoff: faster, fully autonomous execution against direct, traceable human accountability for every individual trade decision.

The earnings call data adds a layer of urgency to that judgment. AI mentions across these six companies' earnings calls nearly doubled in the first quarter of 2026 compared to the fourth quarter of 2025, according to Jefferies data as reported by Investing.com. 

That trajectory suggests none of these companies view their current tier as a permanent resting point. The platforms offering full autonomy today were, in most cases, offering conversational tools or no AI trading product at all a year earlier. 

But the Coinbase Advisor question is the one none of the six companies have actually resolved, and it cuts across every tier. Full autonomy raises a regulatory question. Supervised autonomy raises an operational question. Conversational-only tools raise neither.

FINRA's 2026 Regulatory Oversight Report has already flagged AI agents acting without a human in the loop as a top investor risk, warning that misaligned reward functions could lead an agent to optimize for outcomes that work against the investor it is supposed to serve. 

Coinbase Advisor is now the test case for whether a platform can claim fiduciary duty and disclaim responsibility for that duty's outcomes at the same time. No court has answered that question yet.