Hawk
Cloud-native anti-money-laundering and fraud detection, focused on reducing false positives
Hawk is a financial-crime prevention startup from Munich, Germany that runs a traditional rule engine and machine learning in parallel: rules handle the explicit thresholds required by regulators, models find anomalies rules can't catch, and both results are presented together to investigators.
Features and use scenarios
Its core selling point is explainability — every alert can be traced back to the rule or model feature that triggered it, giving compliance staff a basis when explaining to regulators. The system is designed to be cloud-native, connected via API, adopted faster than traditional packages, with customers mainly European digital banks, payment institutions, and small-to-medium banks, and expanding to North America in recent years.
It suits growing digital banks, payment institutions, and e-money institutions — those whose transaction volume is already too large to review manually but who lack a big bank's budget to maintain a traditional system. If Taiwan's e-payment operators are evaluating an AML-system upgrade, this kind of cloud-native solution is worth comparing prices on.
Main features
- Rule engine and machine learning in parallel
- Explainable alert traceability
- Real-time transaction monitoring
- Sanctions and watchlist screening
- API-first cloud-native architecture
Common uses
- Digital-bank transaction monitoring
- Payment-institution fraud detection
- Alert optimization for existing systems
- Digitizing compliance-investigation processes
Key Features
- Rule engine and machine learning in parallel
- Explainable alert traceability
- Real-time transaction monitoring
- Sanctions and watchlist screening
- API-first cloud-native architecture
Pros
- Adopted faster than traditional packages
- Alerts are explainable, easing regulatory response
- Reducing false positives is a clear product pitch
Cons
- The company is smaller, with limited very-large-institution adoption cases
- Asia-Pacific local support needs confirming
- Cloud deployment is a challenge for some institutions' security policies
Use Cases
- Digital-bank transaction monitoring
- Payment-institution fraud detection
- Alert optimization for existing systems
- Digitizing compliance-investigation processes
Editor's Note
An AML system's biggest hidden cost is manpower — not the money to buy the system, but how many people review false alarms daily. That's the number to use for evaluating benefit.
FAQ
Won't pure machine learning do? Why still need rules?
Regulators require certain scenarios to have explicit, auditable thresholds (e.g., reporting above a specific amount), which only rules can do. Machine learning fills in anomaly patterns beyond the rules; the two complement rather than replace each other.
Can I use it just to reduce my existing system's false positives?
Yes — this is a common adoption path: keep the existing system running, feed alerts in for secondary scoring and ranking, which is lower-risk and easier to persuade internally.
