Q: What makes CipherOwl’s Risk Engine special?

• Unified cross-chain graph. We ingest every major blockchain (EVM, non-EVM, L2s, bridges) into a single, high-fidelity knowledge graph, allowing an address or contract to be tracked seamlessly across networks. An EVM entity therefore has one consistent risk score, no matter which chain it appears on.
• On-demand, organization-specific models. Our risk engine is built for speed and modularity, so each customer can adjust thresholds, weights, and decay factors and receive an updated scorebook in real time.
• Evidence-based, address-level signal. Every risk score comes with a line-item audit trail that shows exactly which hops, counterparties, and on-chain evidence contributed to it. We never hide exposure behind coarse “cluster” labels that inflate false positives; instead, each address (or transaction) is scored on its own merits, so investigators can trace risk back to verifiable data within seconds.
• Future-proof architecture. New chains, data feeds, or regulatory rules plug into the graph as additional nodes and edges, so your risk program evolves in hours—not quarters.

Q: How does CipherOwl guarantee label quality?

• Multi-source pipeline. We blend labels from premium commercial vendors, CipherOwl’s in-house analysts, strategic partners, and (where available) regulatory disclosures to maximize coverage and accuracy. Accroding to back testing,
• Legal-grade evidence chain. Every label carries a verifiable collector audit trail that shows who supplied it, when, and with what confidence - crucial for external audits or legal discovery. Our goal is a “court-ready” standard: each label is versioned, timestamped, and stored with supporting artifacts so it can stand up to regulatory scrutiny.
• Private, customer-specific labels. Institutions can add or override labels unique to their business logic, instantly propagating those changes through their own risk calculations without exposing proprietary intelligence to other users.
• Top quality. According to backtesting results from our partners, CipherOwl's label quality is on par with or exceeds that of the top vendors in the market.

Q: How does CipherOwl leverage AI responsibly?

• Truth over novelty. Large language models are fine-tuned to favor factual accuracy and provenance citations, with creativity parameters intentionally dialed back for compliance workflows.
• Retrieval-augmented responses. Our agents fuse CipherOwl’s on-chain graph with live, external data (court filings, sanctions updates, news) to surface the most recent ground truth while flagging any uncertainty.
• Continuous validation loop. Each AI output is scored against internal heuristics and, when needed, escalated to human analysts—creating a feedback cycle that systematically improves both the model and the knowledge graph.
• Security & privacy baked in. Sensitive customer data never leaves our controlled environment; inference calls to third-party LLM providers are made through zero-knowledge or redacted payloads to preserve confidentiality.

Q: What are the differences between the hosted and on-premises versions of CipherOwl’s Screening Service (ECS)?

CipherOwl offers two deployment flavors for its Edge Compliance Service (ECS), tailored to different operational and compliance needs: