WISHLane operates a layered transaction-monitoring program that screens every contribution for money-laundering and fraud indicators. It works alongside — and does not replace — Stripe's own controls: Stripe Radar fraud scoring on every card charge, and Stripe Connect Express identity verification of every fund recipient. This page summarizes how it works. For the full reference document, use the PDF above.
How it works
Monitoring has two layers that work together, and runs after each successful payment is recorded:
- A deterministic, rule-based engine evaluates each contribution against explicit AML/fraud signals and raises an alert when any signal trips.
- An AI-assisted analysis layer reviews each flagged case in context and produces an independent risk assessment, rationale, and recommended action for a human reviewer.
Monitoring is best-effort and non-blocking — it never interferes with payment processing — and feeds an administrative review queue where a person makes the final decision.
Monitored signals
Each contribution is checked against the following signals. Thresholds are centrally configurable; the descriptions below reflect current defaults.
| Signal | Trips when | Rationale |
|---|---|---|
| Campaign velocity | ≥ 10 contributions to one campaign within 1 hour | Abnormal inflow rate with no organic explanation |
| Donor velocity | ≥ 5 contributions from one registered donor to one campaign within 1 hour | Repeated rapid funding by a single party |
| Structuring | ≥ 3 contributions from one donor sized just below the manual-review threshold | Classic structuring to stay under a review threshold |
| Donor concentration | One donor funds over $1,000 and more than 50% of a campaign's total | A single party cycling funds through a campaign |
A case is rated high severity if the structuring signal fires or two or more signals fire together; otherwise it is medium. High-severity cases automatically pause the campaign so it stops accepting new contributions until an administrator reviews it. Because funds are held until a creator completes Stripe verification, a flagged, unverified creator cannot extract funds while a case is under review.
AI-assisted analysis
Every flagged case is additionally reviewed by an AI analysis layer built on Anthropic's Claude. It receives the campaign context and an aggregated transaction-pattern summary — summary statistics, not raw transaction exports — and returns a structured assessment:
| Field | Values |
|---|---|
| Independent risk rating | low · medium · high · critical |
| Rationale | A concise, case-specific explanation |
| Recommended action | clear · monitor · request info · escalate · hold |
This assessment is decision-support only. It helps a human reviewer triage faster and more consistently — it does not take autonomous action, move money, or resolve cases on its own. It runs only on cases the deterministic rules have already flagged.
Human review & records
All alerts surface in an administrative queue where a designated reviewer can clear a false positive, escalate for further action, or reopen a resolved case — with the rule signals and the AI assessment shown side by side. Every alert is stored with the signals that fired, the severity, the triggering amount, the AI assessment, and the identity of the reviewer who resolved it, producing an auditable trail.
Where it fits
| Layer | Provided by | Purpose |
|---|---|---|
| Card fraud scoring | Stripe Radar | Real-time fraud risk on every card charge |
| Recipient KYC / identity | Stripe Connect Express | Identity, representative, and bank verification before any payout |
| Contribution limits | WISHLane | Per-transaction, per-donor, and per-campaign caps enforced before payment |
| Held funds | WISHLane | Funds held in the platform balance until the recipient is verified |
| Transaction monitoring | WISHLane | The rule-based + AI-assisted program described on this page |
This page describes controls implemented on the WISHLane platform. Thresholds and figures are centrally configurable and current as of the date above. The AI-assisted analysis layer provides decision support to human reviewers and does not act autonomously.
Policies & documents
WISHLane's compliance policies are available as PDFs, including a written commitment to an independent review of its program at least annually: