Mozn

Two model families carry FOCAL. Screening runs on the bilingual transformer Mozn built for Arabic names, which reads each name together with its context fields. Monitoring layers anomaly detection, ensemble behavioral models and supervised false positive reduction over a configurable rules library, and an investigation agent assembles each case before an analyst opens it. Around those models, the rules engine and case manager are conventional.

Agents prepare and people decide, by Mozn's design. The investigation agent gathers evidence, resolves name matches, drafts the narrative and hands over a recommendation with a confidence level and its reasoning, and the decision to file a SAR remains with a person. Screening clears low risk matches automatically under each client's own rules. Mozn says the agent can also close low risk alerts unaided, but not where that line sits or who reviews those closures.

Rule simulation lets a team test rules and thresholds on data before they go live, every case keeps a full audit trail, and Mozn cites lineage and versioning for its models. The agent explains itself in detail, showing which data points were weighted, which matches were shown or set aside, which thresholds applied and the source record behind each. Mozn gives no accuracy, precision or recall figures for the screening and monitoring models, and cites no independent validation.

HyperPay, a Riyadh payment services provider with about 700,000 users, cut false positives by 45 percent and average alert handling time by 35 percent, with monitoring rules mapped to SAMA's payment typologies. Its chief compliance officer credits Mozn with getting the firm live on time under SAMA's AML and CFT deadlines. An unnamed Saudi bank founded in the 1970s cut fraud losses by 38 percent working with Mozn's forward deployed team.

A Riyadh digital wallet, also unnamed, cut fraud by more than 90 percent with device intelligence. Mobily Pay put customer risk ratings in place in under 30 days, and Aseel, a crowd investing platform, cut onboarding time by more than 87 percent. Mozn credits three customers with an identical set of results, so those figures say little about any one deployment. Its figures across all screening customers, 85 percent fewer false positives and cases resolved 70 percent faster, name no institution.

FOCAL is sold on sovereignty. Mozn says customers keep control of the models, agents and workflows, and FOCAL can run on premises inside the institution, so data need not leave it. Investigator outcomes feed back into detection, and the agent works against each institution's own thresholds and case history. For the hosted service, Mozn does not say whether learning is pooled across customers or feeds the shared screening model, or which model the agent runs on.

Saudi Arabia's Personal Data Protection Law and GDPR are the regimes Mozn says it meets, and it lists a PDPL accreditation from SDAIA, the Saudi data authority. A data processing agreement, a data protection impact assessment, a breach notification policy and a data privacy officer contact sit in its trust center, released on request. Retention periods and subprocessors for FOCAL are not public, and Mozn's undated privacy notices cover advertising rather than the product.

Mozn reports ISO 27001 certification and an EY audit against the Saudi National Cybersecurity Authority's 2024 Essential and Cloud Cybersecurity Controls, though no SOC 2 report. A SafeBase trust center holds more than a dozen security policies, a penetration test report, a vulnerability assessment, a network diagram, the master services agreement and the SLA, most of it under NDA. Data is encrypted with TDE at rest and TLS 1.3 in transit, and a managed detection and response provider watches operations around the clock.

FOCAL is built around SAMA's requirements, from its counter fraud framework to its AML and CFT rules. Mobily Pay rates each customer's fraud risk on FOCAL to meet SAMA's rule on customer level assessment, and HyperPay's monitoring rules map to SAMA's payment typologies. Mozn points to no regulator sandbox or supervised assessment of FOCAL itself, so the product's standing rests with the institutions that run it.

Arabic name matching is the core of the product, and name matching is where screening errors tend to differ by nationality, script and naming custom. Mozn gives its false positive cuts only in aggregate, with no split by name origin or population and no fairness testing. It cites alignment with SDAIA guidance, ISO 42001 and the NIST AI risk framework without certification under any of them, and does not say who assessed its stated compliance with GCC AI requirements.

Mozn offers no public warranty, indemnity or accuracy level for FOCAL's output, and its master services agreement and SLA are shared only on request. Inside the product, every case keeps an audit trail and reasoning that traces to source records, so an analyst can see and reverse what the agent decided.

Core matching runs on Mozn's own transformer, built on proprietary Arabic language work, so it does not depend on an outside model provider. Mozn does not name the model behind the agent's summaries and narratives, the sources behind the more than 1,300 lists built into screening, or its subprocessors. Transaction screening can also run on Dow Jones or World Check data that a customer licenses.

One API connects FOCAL to core banking, onboarding, case management and outside data, though Mozn names no specific core banking system. Transaction screening reads SWIFT MT and MX, ISO 15022 and ISO 20022 messages across SWIFT, real time gross settlement, ACH and instant payment rails. At one digital lender, onboarding connects to Yakeen, the Saudi national identity service, and approves or rejects applicants automatically against set criteria. Saudi integrators NourNet, MDS and Ejada became deployment partners in September 2026.

Customers choose cloud, on premises or hybrid deployment, and Mozn says the customer screening product can also run in country without additional infrastructure. The hosted service is listed on Google Cloud Marketplace, with the local host Edarat as a residency partner, though Mozn does not name its regions. Mozn's wider platform installs in local infrastructure, private cloud or fully air gapped environments, which is how it sells to Saudi banks and government, where residency is a legal requirement.

A free 30 day demo opens the sale, and FOCAL is then priced by quote, with no list price or unit of sale. A startup program offers a discount of undisclosed size to firms with less than $1 million in funding that commit to a 12 month contract. Deployment, rule tuning and periodic system assessments come as professional services, delivered by more than 120 forward deployed engineers who work alongside a customer's fraud operations, MLRO office and model risk staff.

Banks, payment firms, wallets, digital lenders and fintechs are the core buyers, with insurers and other nonbank institutions also served. Al Rajhi Bank, stc Bank, D360 Bank and Abdul Latif Jameel Finance have signed on, alongside HyperPay, Mobily Pay and Aseel. Across financial services and the public sector, Mozn counts more than 150 customers.

Coverage runs from customer due diligence and sanctions screening through payment screening, transaction monitoring and enterprise fraud, where device intelligence and behavioral monitoring sit beside transaction fraud rules. The footprint is the Gulf, with offices in Riyadh and Dubai and named customers concentrated in Saudi Arabia.