Fraud prevention
Fraud prevention built for funded-trader economics.
Prop firm fraud is not generic e-commerce fraud. The economics are specific: an evaluation fee is small, a funded payout is not, and the profitable attack is to buy many accounts cheaply and engineer a payout from at least one of them. That produces patterns — account farming, coordinated trading between accounts, hedged pairs across firms, identity reuse behind fresh payment instruments — that a general fraud tool is not looking for.
Nexus treats fraud as three connected risk surfaces with human review at the end, and every flag carries an identifier so a decision can be explained later.
A complete fraud intelligence system
Identity risk
Device and browser fingerprinting, identity verification signals and reuse detection across accounts that present as unrelated.
Trading risk
Coordinated trading, hedged pairs, mirrored entries and the statistical fingerprints of accounts being traded as one book.
Financial risk
Payment instrument reuse, chargeback history, payout destination patterns and the mismatches between who paid and who withdraws.
Every flag carries a Risk ID
A flag without evidence is an accusation. Each signal Nexus raises is recorded with an identifier, the evidence behind it and the time it was produced, so an operator reviewing a payout can see precisely why the system is uneasy — and so a trader who is wrongly flagged can be cleared with a record rather than a shrug.
This matters commercially as well as ethically. Payout denial is the highest-friction action a prop firm takes, and it should be defensible.
A fraud agent working the queue in the background
Fraud review does not scale by hiring reviewers, because the evidence a case needs is scattered across trading, payments, devices and payout history, and assembling it is most of the work. The Nexus fraud agent runs continuously against open reviews and does that assembly before a human opens the case.
It is built evidence-first, and the ordering matters. Deterministic collectors own the facts — payout request IDs, masked account IDs, fill and order IDs, reconstructed trade windows, payout-period PnL attribution, device and fingerprint links, shared payment instruments, affiliate relationships, duplicate-identity flags and the relevant policy text. The model never sources a fact; it reads what the collectors gathered.
The agent then scores the case, explains its reasoning, and — the part reviewers actually value — names what evidence is missing. It recommends one of a fixed set of outcomes: monitor, manual review, hold payout, reject payout review, or blacklist review.
What the agent does, and what it is not allowed to do
Works cases before you open them
Assessments run in the background as evidence changes, so a reviewer opening a payout sees an assembled case rather than a starting point.
Records only when its view moves
An assessment is written when the risk score, recommendation or summary actually changes. Logging every recomputation would bury the moments the picture genuinely shifted.
Answers questions on the case
Reviewers can interrogate the assessment in an append-only transcript alongside it. The question is recorded before the model is called, so the record reflects what was asked even when the agent cannot answer.
Refuses to guess
Chat is grounded strictly in the stored evidence for that case. Where the evidence does not settle a question, the agent says so and names the gap instead of inferring.
Never decides
Payout rejection, profit removal, breach enforcement, suspension, blacklisting and cross-firm sharing are human actions. The agent recommends; an operator decides, with attribution.
Keeps an indefinite record
The transcript is append-only and never deleted, so what the agent advised before an enforcement action remains auditable long after the fact.
Catch the ring, not just the individual
Fingerprint-backed account uniqueness
Detect one person operating many accounts even when identity documents, emails and payment methods differ.
Cross-firm hedge alerts
Hedging across two firms is invisible to either one alone. Network participation is what makes the pattern detectable at all.
Blacklist once, protect the network
A confirmed bad actor blacklisted at one participating firm becomes a signal for the rest, rather than a lesson each firm pays to learn.
Block the person, not just the account
Enforcement targets the underlying actor, so closing an account does not simply move the problem to the next signup.
You do not have to run all of Nexus to join
Fraud intelligence is more valuable the more firms contribute to it, so network access is available on its own as well as included with the platform. A firm running someone else's trading stack can still participate.
Commercial terms, what participation requires and what is shared are scoped directly — see engagement models or raise it on a call.
Keep reading
Back to hub
Platform
One platform connecting the branded trader experience with CRM, risk, fraud prevention, payouts and support so every team works from the same account context.
ReadRelated
Operations CRM
Run customer lifecycle, communications, payout review, support and provider-neutral trading integrations from a single operator workspace built for prop firms.
ReadRelated
Payout operations
Designing payout operations traders trust: eligibility rules, review queues, fraud checks, evidence retention and the controls auditors expect to see.
ReadNext step
Contact
Talk with Nexus about launching, migrating or building a funded-trader platform, and get a scoped view of deployment, integrations and operating support.
ReadTalk about the fraud you are actually seeing.
Bring the patterns that have cost you money. Those are the ones worth testing the network against.