An AML score is a triage tool, not a cleanliness certificate. It condenses the evidence around a Bitcoin wallet into a 0 to 100 result so a reviewer can decide which transaction needs closer inspection.
Evidence enters the model first
The assessment maps the BTC transaction graph to attributed entities and risk categories. Direct interactions carry different weight from indirect paths. The model also needs the direction and scale of the relevant flow.
Bitcoin signals need their own context
Bitcoin is the original proof-of-work blockchain, the most widely tracked by OFAC and compliance teams worldwide. Evidence from that network sits beside sanctions, darknet, mixer, ransomware, scam, and stolen-funds findings. A category label contributes to the result; it does not explain the result by itself.
Scoring policy belongs to the organization
Two firms can observe the same on-chain evidence and choose different review thresholds because their customers, products, and regulatory duties differ. Keep the raw findings separate from the policy decision. That separation lets a team change category weights without rewriting transaction history or pretending yesterday's score was a permanent fact.
Good calibration compares model bands with reviewed outcomes and watches for categories that flood the queue without changing decisions. The goal is useful prioritization, not a dramatic number. A threshold that sends every wallet to manual review has failed even if its mathematics looks sophisticated.
Analysts need override reasons as structured evidence: confirmed exchange ownership, verified source of funds, stale attribution, or a genuine severe link missed by the aggregate. Recording those reasons improves later calibration and gives auditors a trail from raw graph evidence to the final disposition.
The final action is categorical even when the input is numeric: process, review, restrict, or report under a defined policy. Writing that mapping down prevents staff from improvising a different meaning for the same band on each case.
Coverage limits belong in the result as well. Incomplete attribution or limited visibility reduces what the engine can observe. A low total therefore means low detected risk in the available evidence, never universal clearance of the wallet.
Read the four bands as review priorities
- 0–25, low: the screen found no significant flags in the available data.
- 26–59, medium: review indirect or limited exposure and collect context.
- 60–84, high: investigate the material finding before processing funds.
- 85–100, critical: stop the routine flow and escalate direct severe exposure.
Keep the evidence beside the number
Run the wallet above, open the category breakdown, and document the transactions behind material flags. Your policy, jurisdiction, customer context, and source-of-funds evidence determine the final action. This information is not legal or financial advice.