A counterparty who learns your crude position before the settlement window moves the market against you. A trader who carries a P&L view out the door to a competing shop carries the edge with it. On an energy or commodities desk the live book is worth more than most of the physical barrels behind it, and it never sits still. It moves from the trade-capture system to the risk engine, out to the ETRM, into a spreadsheet on the desk, across to clearing and counterparties, and now into the AI tools analysts point at the book. The desk granted every one of those paths on purpose. It cannot trade without them.
So the only question your security stack knows how to ask, "was this move permitted," comes back yes. Every time. The move that walks the firm out the door comes back yes too.
The breach looks like the desk working
A trader two weeks from his last day runs the settled-trade export he runs every Friday. Nothing blocks it, because the export is his job. An analyst syncs the position book to a personal cloud folder to work from home, down a sharing path IT signed off on two years ago. A counterparty integration that normally pushes confirmations starts pulling a wider slice of the book than the contract covers, on credentials that are entirely valid.
Read one at a time, each is a clean, sanctioned action. The danger lives in the shape across them. Which identity. Which book. Which destination. What volume, at what hour, set against the months of history that define normal for that desk. A tool that evaluates moves one at a time cannot see a shape that only exists across many. That shape is the breach.
Your stack was built for a different problem
DLP and email security catch the clumsy egress: the position file attached to a personal Gmail, the keyword that trips a policy. They match content against rules, so the move that rides a channel you sanctioned for a real reason walks straight through. The ETRM logs who touched what, but a log is a record you read after the fact, not a live read on whether the touch fits. Endpoint tools hunt malware on the workstation. Network controls decide who can reach what. Not one of them resolves a data movement to the trader and the job behind it and asks whether this read, at this hour, fits how this book has moved for the last six months.
That leaves the desk with two bad options. Predictive tools guess ahead of the move, fire on the legitimate end-of-quarter export, and teach the desk to swipe the alerts away. Forensics names what left after the position is already in a rival's hands and the only thing left to write is the disclosure. The live middle, watching the movement as it happens, is the part nobody shipped.
Runtime governance, on the movement itself
Hilt is that middle. One lightweight collector watches data movement at the kernel, metadata only by default, off the path. It runs single-tenant inside your own cloud (AWS, GCP, Azure, or Ali Cloud) at roughly 0.1% of one core and 4 to 8 MB of memory per host. It does not sit between your trade systems and where the data goes. It never blocks, drops, or alters a move. It watches.
Each move resolves to a probabilistic, source-dependent identity and the job behind it, then gets scored against what that identity normally does with that data. Metadata carries the shape of a move, so Hilt reads that a position file is leaving in a way that does not fit without reading the file. Content-aware inspection is there when a case warrants it. It is never the price of admission.
Go back to the departing trader. The export channel is approved, so the rule-based tool stays quiet. Hilt sees the deviation across layers at once: the job is off for his identity this week, the read spans more historical trade data than his role normally touches, and the destination volume runs high for that path even though the channel is sanctioned. Any one signal is noise. Together they are a pattern, and a pattern is a case, not an alert. It arrives written, naming the identity, the book, the destination, and why the shape is wrong, while there is still time to act instead of after.
When a case crosses the line, Hilt isolates that host at the network from the control plane, so the move stops without anything ever having sat inline. The desk keeps trading. The one host carrying the anomalous movement is the only thing that goes quiet.
What a commodities CISO actually buys
Your stack covers most of the surface and is good at what it covers. DLP catches the obvious egress. The ETRM enforces entitlements and keeps the record. Endpoint and network tools handle malware and access. Hilt replaces none of it. It adds the one read none of them was designed to give: a live, identity-resolved view of the pattern of data movement across moves that are all individually permitted.
Three things make that read fit a desk where heavier controls have not. It stays off the path, so it adds no latency and never stands between the trade system and the market. It is metadata-only by default, so the live book, the most sensitive thing you own, does not have to be read by your security tooling to be watched. And events never leave your account, which is the whole game when the data in question is the position and the regulators, the counterparties, and the competition all want to know where it went.
The blind spot is not unique to trading. Every firm moving valuable data on access it granted on purpose carries it. The desk just makes the stakes legible, because the thing that moves is the position, and the position is the money.
If your controls cannot answer what this book's data did last week, resolved to the trader and the job behind it and scored against how it normally moves, that is the gap. Thirty minutes, engineer to engineer, is enough to walk your position-data paths and see where the pattern would surface.