Smarbl

SmartReg's rules, regulatory data model and submission are deterministic and would survive without models, while RegPRISM, the regulatory intelligence product, would not. Extracting an obligation from regulatory text and interpreting it is model work, so the newer product depends on AI while the flagship does not.

Validations, reconciliations and data quality checks run before a return is built, and approvals, errors and rejections move through a customizable workflow with oversight dashboards. Every control governs SmartReg, while RegPRISM's obligation extraction has no described reviewer step or confidence threshold.

Obligation extraction is measurable, captured from the source text or not, yet no accuracy figure, benchmark or validation method is published for RegPRISM. Evidencing correctness is clearly understood on the deterministic side, yet none of it is published for the model side.

United Arab Bank co developed RegPRISM and announced its July 2025 launch with its chief executive quoted. The homepage names no customer and leans on Chartis RiskTech100 and IDC recognition, which describes standing rather than results.

RegPRISM was shaped in part by a rival bank's compliance work, so a competing buyer gets something another institution helped build. What the bank contributed, whether it keeps an interest and whether one customer's obligation mappings reach another are not described.

The homepage and SmartReg page carry no privacy statement, retention rule or data handling commitment. Its data is mainly institutional supervisory material, though anti money laundering reporting identifies individuals under investigation.

OAuth2 and SAML authentication and data encryption are described, with no certification, audit report or trust center named. Supervised banks open their third party risk reviews with exactly that question.

Registration in Abu Dhabi Global Market, a financial free zone with its own regulator, is checkable, and pre configured returns must satisfy each supervisor's schema, submitted in formats including XBRL. No supervisor has examined or admitted RegPRISM.

What RegPRISM surfaces determines what a compliance team believes the law requires, so coverage is the exposure rather than protected class discrimination. Coverage across regulators is untested in public, and no governance framework or independent assessment sits behind the AI driven label.

Lineage, audit trails and logs let an institution trace how a return was built, but no accuracy warranty, service commitment or remedy is published. A filed return goes to a central bank in the institution's name, so a faulty template becomes the bank's supervisory problem, and RegPRISM carries the mirror risk of an obligation never surfaced.

Provider, base model and hosting for RegPRISM are unnamed, and SmartReg is described only as built on open source technology. Choosing on premises for residency would mean little if RegPRISM calls an external model endpoint, and nothing says whether it does.

Data from core banking, risk and financial crime systems is ingested into the regulatory data model, and returns go out in regulator formats. Not one core banking, ledger or risk platform is named, and no connector catalog or developer documentation is published.

Cloud, on premises and hybrid models are stated for SmartReg, with on premises named as the answer for data residency concerns. Hosting region and tenancy model are not listed, and cloud providers are described rather than named.

A demo request is the route to a number, with no rate card, tier ladder or billing basis published. Zero costs and lower capital spending through pre built reports give the direction of the argument but none of its magnitude.

Capital, liquidity and large exposure returns share one data model with anti money laundering reporting, covering much of what a bank files. Pre built returns exist for selected regulators, the named customer is in the United Arab Emirates, and global reach is a stated strategy rather than a footprint.