The financial-crime engine for banks,
fintechs and payment companies.

See the engine decide, in real time

Every transaction is scored, every party screened, every high-risk decision turned into a case. Allow, hold or block, with the reason attached. The stream below is illustrative.

Financial crime is universal. Your controls should be too.

Fraud rings, mule networks, sanctioned counterparties and politically exposed persons move the same way in Lagos, Nairobi, Johannesburg, London and Dubai. Most tools are trained on one region's crime and localised slowly, in a foreign currency. Omniguard starts from a single global engine and adapts down to each market, so an institution anywhere gets first-class screening, monitoring and reporting from day one.

One decision, the full obligation

Every transaction, every counterparty and every report passes through the same explainable engine, and every decision is recorded for the examiner.

• Real-time transaction scoring. Every payment is scored in milliseconds by tuned rules and two self-training models per product line. Allow, review or block, with the contributing reasons attached.
• Sanctions and PEP screening. A single consolidated watchlist of OFAC, EU, UN and UK sanctions plus global politically exposed persons and national enforcement records. More than 60,000 entities, fuzzy and alias-aware, tuned for local names and honorifics.
• Adverse-media screening. Screens a name against current negative news, fraud, laundering, corruption and enforcement coverage, and flags corroborated, evidence-linked hits for enhanced due diligence.
• Identity verification. Verify national IDs and passports on the Nemesis identity network during onboarding or a live transaction, and fold the result into the same risk score.
• Regulatory reporting. Suspicious-transaction reports generated in goAML, with a valid predicate offence and red-flag indicator already attached, so filings pass validation on the first submission.
• Cases, audit trail and API. High-risk decisions open cases with SLA timers, every check is logged for examination, and any control is callable as a standalone API from your own systems.

The AI advantage: two models that learn your business

Rules catch what you already know to look for. Omniguard adds machine learning that learns what you do not: a supervised model and an anomaly-detection model, trained per product line on your own traffic, not a shared, one-size-fits-all model.

Both models are built per function, retrain automatically as data accumulates, and pass promotion gates with rollback before they ever move a score. Their output fuses with the rules into one explainable number:

Built for the AML and CFT typologies that actually move money

Anti-money-laundering and counter-terrorist-financing are not one rule. Omniguard ships tuned detection for the patterns regulators and FATF care about, and every hit lands as an explainable case with a predicate offence and indicator ready to file.

Reporting the way financial intelligence units actually read it

goAML is the reporting platform built by the United Nations Office on Drugs and Crime and used by financial intelligence units across Africa and around the world. Modern goAML validation rejects any suspicious-transaction report that does not carry a valid predicate offence and a recognised indicator. Omniguard attaches both, from the controlled vocabulary, before the report is ever exported, so your filing is accepted the first time instead of bouncing back.

The compliance obligations it satisfies

Omniguard maps to the controls examiners and FATF assessors look for, and produces the evidence to prove them, so a compliance officer and a supervisor can walk the same checklist.

• Risk-based approach (FATF R.1). Every customer and transaction carries a live, explainable risk score, tuned per product line rather than a static rulebook.
• Customer due diligence and EDD (FATF R.10, R.12). Identity verification at onboarding, and a PEP or adverse-media hit automatically escalates the party to enhanced due diligence.
• Targeted financial sanctions (FATF R.6). Real-time, alias-aware screening of every party against OFAC, EU, UN and UK lists, before the funds move.
• Wire transfer information (FATF R.16). Originator and beneficiary details are part of the data contract and screened on every transfer, supporting travel-rule obligations.
• Suspicious transaction reporting (FATF R.20). STRs generated in goAML with a valid predicate offence and indicator, so filings are accepted on the first submission.
• Record-keeping and auditability (FATF R.11). Every decision, its contributing signals and its screening evidence are retained for examination.
• Regional mandates.Reference-built to Nigeria's CBN Baseline Standards for Automated AML Solutions and NFIU goAML, and configurable to the FIU format and sanctions regime of any market you operate in.

Local rails, global reach

The engine is global. The last mile is local, and Omniguard treats that as a feature, not a rebuild.

• Any currency and channel. Native scoring, thresholds and reporting across instant transfers, cards, wallets and mobile money, in the currency you actually settle in.
• Any national enforcement list.Global sanctions and PEPs come built in; a country's own enforcement and wanted records plug in alongside them. Nigeria's EFCC data ships as a worked example.
• Any identity system. Verify against the national rails that matter in your market. BVN, NIN and international passport are supported today, with more identity sources on the same interface.
• Names as your customers carry them. Matching that understands local honorifics, spelling and word order, so real hits are not missed and clean customers are not stopped.

Who runs Omniguard

One engine, many front doors. Each institution turns on the controls it needs and calls the rest as an API.

Proven at the strictest bar

Omniguard was engineered and proven against Nigeria's CBN Baseline Standards for Automated AML Solutions and the NFIU goAML requirements, among the most demanding automated-compliance mandates any regulator has published: real-time screening, risk-based monitoring, first-pass suspicious-transaction reporting, model governance and full auditability. Clearing that bar is the proof, not the boundary. The same engine drops into any institution that needs fraud and AML controls it can stand behind in an examination.

See the full regulatory mapping in a supervised pilot →

Call it from your own stack

Score a transaction, or run any single check, over one authenticated API. Screening and identity work standalone, so you can use Omniguard as a verification service without wiring a full monitoring function:

POST /v1/omniguard/score        # full fraud + AML decision
POST /v1/omniguard/verify       # one check: sanctions | pep | adverse_media | bvn | nin | passport

{ "amount": 4500000, "currency": "NGN", "channel": "transfer",
  "beneficiary_name": "…", "customer_ref": "cust_10293" }

→ { "verdict": "review", "overall_score": 83,
    "reasons": ["PEP / enforcement match", "velocity 8x baseline"],
    "predicate": "fraud", "indicator": "I-04" }

Every response is explainable and metered against your plan. A missing required field is reported, never silently ignored.

Honest by construction

• Never fabricates a pass. If a check cannot run, it says so and holds for review. It does not invent a clean result.
• Fails open on infrastructure, closed on risk. An internal error never silently blocks a legitimate payment, and a genuine hit is never released.
• Every decision is explainable and retained. Contributing signals, scores and screening evidence are recorded for the examiner.
• Deploys as a decision layer. One API in front of your core and payment systems. No rip and replace.