AMLBot

Risk scores average the conditional scores attached to an address's on chain links, which is aggregation rather than inference. The newer AI Tracer applies AI to fund tracing, but strip the models from the core screening product and a risk data aggregator with alerting and an investigation interface still stands.

Each customer sets its own risk thresholds, and alerts land with a person who reviews, prioritizes or dismisses them in a dashboard; published guidance lists what an audit trail must hold, down to the decision taken and who approved it. Nowhere does it say whether automatic acceptance or rejection can be configured or what the defaults do.

A 99.5 percent data accuracy claim arrives with no definition of accuracy and no sample, period or method behind it. Recall against a known illicit set, a validation method and notice of coverage changes are absent as well.

Eight customers are named, BTSE, ChangeNow and Banana Gun among them, alongside aggregate claims of more than $100 million in risky funds detected and more than 2,500 investigations in 2025 and 2026. None of those customers carries a figure, and the customer count itself reads 300 on the homepage and 1,000 on the about page.

Resultant data survives termination under the user agreement while other licensee data is erased, and improving the service is a listed use of personal data. Whether one customer's screening queries or verified identities shape the risk scores others receive, or can be withheld, is left unstated.

Screening and identity verification fall inside the privacy policy, which names Safelement Limited in Hong Kong and Safe3 UAB in Lithuania as controllers by customer residence, covers biometric data from document checks and keeps inquiry data for up to three years. Transfers out of the European Economic Area rest on model contract clauses, and processors appear only by category, with Google Analytics the lone name.

ISO 27001:2022 and ISO 9001:2015 certifications appear on the certifications page as badges, with no certificate document, issuing body or scope, and the user agreement adds a 99 percent monthly availability commitment. Neither a SOC report nor a trust center is offered, leaving open whether the identity verification store sits inside the certified scope.

Its material works through named regimes, FATF recommendations and European anti money laundering standards among them, builds originator and beneficiary transfer into the workflow, and sells consulting on obtaining crypto licenses. For itself, though, it states no authorization, registration or supervisory relationship, the usual posture for a technology supplier.

Taint travels along on chain links, so someone who received funds from a tainted counterparty carries part of it through no act of their own, and AMLBot concedes screening cannot guarantee against exchange holds. It publishes no false positive rate and no analysis of how scores fall by jurisdiction or by services common in particular regions.

Under the user agreement the service comes as is, liability is capped at fees paid, the licensee is responsible for decisions based on outputs and must indemnify the company. A person whose funds are refused on an AMLBot score is not its customer and has no published route to contest it.

Three data sources feed the risk scores and none is named, though in blockchain analytics the attribution provider decides chain coverage and how fast new illicit services are recognized. AI Tracer names no model provider either, and the identity verification partner handling document images is given only by category.

Reference documentation covers the transaction monitoring API, AMLBot says it handles integration itself so the customer's developers stay on their own work, and identity verification goes live in a day. Coverage spans multiple blockchains, bridges and stablecoins, yet the homepage and about page name no exchange, custody platform or wallet provider as a ready connector.

Identity checks collect document images, facial biometrics and liveness captures across 240 countries, yet the hosted service names no provider, region or residency commitment for that material. Several of its markets regulate where such data may be processed.

One free offer is concrete: an address can be verified once and then monitored on a continuing basis at no charge. Paid use has no published unit, whether screened address, monitored wallet or seat, and every other route runs through a conversation with the team.

Exchanges, wallets and virtual asset service providers are all addressed, with material pitched at teams too small to build compliance in house, and named customers run from wallets such as Trustee to protocols such as Railgun. Reach is claimed across 25 jurisdictions, yet no bank, payment institution or custodian appears among the names.