There's a question that keeps coming up in dealing desk conversations lately, usually phrased something like this: "We started seeing patterns we didn't recognise. Same timing, same size, same reaction to price — across accounts that have nothing to do with each other."It's not a new problem. Coordinated behaviour, latency arbitrage, group exploitation — brokers have been managing these for years. But something shifted in the last few months. The patterns are cleaner. More consistent. Less human.What shifted is that in many cases, they aren't human.The Infrastructure Just ChangedIn May 2026, Spotware opened cTrader to AI agents through Model Context Protocol — letting external AI tools place trades, manage positions, and control charts through natural language prompts. TraderEvolution had done the same in January. In March, two engineers at Revolut built a working market-making system in roughly half an hour using AI tools. Finance Magnates covered all three. Nobody in the industry missed it.What's less discussed is what this means on the other side of those trades — at the broker's risk desk.The trading infrastructure has just opened itself to a class of participants that behaves fundamentally differently from retail clients. AI agents don't hesitate. They don't get distracted, don't deviate from their logic, and don't take breaks during the London open. They execute consistently, at scale, with the same parameters every time. In normal conditions, that looks unremarkable. Under specific market conditions — a spike, a spread widening, a liquidity gap — it looks like something a dealing desk needs to have seen coming.Read more: ThinkMarkets Launches MCP Server; “AI Can Execute Trades, but Not Access Funds”What Human Intuition Was Built ForRisk management in retail brokerage was designed around human behaviour. Humans are inconsistent. They overtrade after losses, freeze during volatility, cluster around round numbers, and react emotionally to news. That inconsistency is, paradoxically, part of what makes a large retail book manageable. The noise averages out.An AI agent has none of that noise. It does exactly what it was built to do, every time, without variation. Which means if it was built to exploit a specific condition — a price feed lag, a spread pattern, a session gap — it will exploit it completely, consistently, and at whatever scale the account allows.The FM Intelligence data from Q1 2026 puts context around this. Active CFD accounts hit 7.4 million — up 42% year on year. Average monthly volume per 1,000 active accounts reached $4.3 billion, up 27% from Q1 2025. The spread between the most and least active brokers in the cohort was 17-fold. Some of that volume is human traders getting more active. Some of it is something else, and the dealing desk, looking at raw numbers, can't immediately tell which is which.The Detection ProblemHere's what makes this genuinely hard. An AI agent operating a legitimate strategy looks, at the surface level, like a disciplined human trader. Consistent sizing. Consistent timing. Consistent reaction to market conditions. The signals that traditionally flag suspicious behaviour — unusual patterns, erratic execution, timing clusters — are exactly what a well-designed agent will not produce.Related: Claude Powers Nine of Ten Broker AI Agents That Now Trade Live AccountsThe flags that matter are different. Not "this looks strange" but "this is too consistent." Not "this account is correlated with others" but "this account's behaviour changes in a mathematically predictable way when a specific condition occurs." Not volume or frequency, but the relationship between market microstructure and execution timing — and whether that relationship holds across accounts that shouldn't have any connection.This is pattern recognition at a level that's genuinely difficult to do manually. A dealer looking at a screen sees an account that's performing well, trading normal sizes, not triggering any obvious alerts. The problem isn't visible in any single account. It's visible in the relationship between accounts — and between those accounts and specific market conditions — over time.That's exactly the kind of picture that requires aggregated, real-time visibility across the full book, not account-by-account review. The dealer needs to be able to see it. And then the dealer needs to decide what to do about it — adjust spreads, flag the group, change execution conditions, escalate. Visibility is the precondition for the decision. Without it, the decision comes too late or not at all.The Speed AsymmetryThere's another dimension worth naming directly. AI agents operate at machine speed. A dealer reviewing alerts, cross-referencing accounts, assessing exposure, and forming a judgment operates at human speed. That gap exists in traditional algorithmic trading too, but it narrows significantly when the agent is running through a retail platform with retail execution conditions — because the broker's own infrastructure introduces latency that partially equalises things.What AI agent connectivity via MCP changes is that the agent is now sitting much closer to the execution layer. The interface friction that used to slow things down is reduced by design. The broker's advantage — that retail conditions inherently limit how fast an external actor can move — shrinks.This doesn't mean the broker loses. It means the broker needs to be faster at recognising what's happening, so that the human making the decision has enough time to actually make it. Early visibility isn't a nice-to-have in this environment. It's the margin between a considered response and a reactive one.What Changes for Risk DesksThe practical implication isn't to panic. Most brokers won't suddenly find their books overrun by sophisticated AI agents tomorrow. The shift is more gradual — more algorithmic behaviour, more consistent patterns, more edge cases that don't fit traditional toxicity profiles.But the direction is clear, and the dealing desks that handle it well will be the ones that update their mental model of what "suspicious" looks like before the volume becomes a problem, not after.That means looking beyond individual account flags to cross-account pattern analysis. It means paying attention to the relationship between execution behaviour and specific market microstructure events — not just whether something happened, but when it happened relative to what else was happening in the market. It means treating "too consistent" as a signal, not a reassurance.None of this replaces the dealer's judgment. The dealer still decides. But the dealer can only decide well when the picture in front of them reflects what's actually happening — and right now, what's actually happening is changing faster than most risk setups were built to track.This article was written by Marina Koltsova at www.financemagnates.com.