A trader meant to sell $58 million. The system built a $444 billion order — and $189 billion slipped past the controls. This was Citigroup, May 2022. The pre-trade controls caught $255 billion of the basket. But the rest reached the trading algorithm, which broke it into sell orders and pushed $1.4 billion into Europe before anyone could cancel — setting off a flash crash. The roughly $92 million fine, across BaFin and UK regulators, wasn't really about the fat finger. Citi could show the orders. What they couldn't show was why their controls let $189 billion through. That's the shift we keep running into — across SEC, MiFID II, and the EU AI Act. Regulators have stopped asking "do you have controls?" They're asking "can you reconstruct every decision your algorithm made?" The FCA's 2025 review of ten trading firms found most couldn't even produce a current inventory of which algorithms they run. The catch: the surveillance platforms built for this — NICE Actimize, Nasdaq — run $1–5M+ a year and target Tier 1 institutions, leaving mid-size banks facing the exact same rules without the budget or the staff to run them. So we build it the other way: explainability-first compliance systems where every algorithmic decision carries its own auditable reasoning chain, not just a log of what happened. If an examiner asked you to walk through what your algorithm did at 9:47 one morning — the reasoning, not the order ticket — could you? #AlgoTrading #TradeSurveillance #RegTech