Mashreq deploys generative AI fraud platform with Clari5 in UAE
Mashreq, one of the UAE's oldest and largest financial institutions, has gone live with a generative AI layer inside its fraud risk management platform, built in collaboration with Clari5, the financial crime management unit of India's Perfios group. The deployment, announced at Mashreq's Innovation Hub in Dubai, positions the bank as an early adopter of enterprise-grade generative AI in the Gulf's regulated banking sector, and raises broader questions about how compliance-sensitive industries elsewhere will reconcile AI capability with auditability.
The platform is designed to accelerate fraud investigations by synthesising disparate data sources, surfacing patterns for human analysts rather than acting autonomously. Clari5 says the system runs on-premises inside Mashreq's own infrastructure, keeping decision logic within the bank's governance perimeter. That design choice is deliberate: in an era when regulators across the EU, UK, and GCC are still finalising AI liability frameworks, the ability to demonstrate an auditable, explainable model is commercially as well as legally valuable.
Governance as a competitive asset
Rivi Varghese, co-founder and CEO of Clari5, framed the deployment in terms that will resonate far beyond fraud teams: "Banks don't need AI that acts on its own. They need AI they can govern by their own rules." That framing captures a tension running through every regulated industry attempting to adopt large language models, from insurance underwriting to pharmaceutical trial monitoring, where the cost of an unexplainable decision can be regulatory censure, not merely a bad outcome.
Clari5's parent, Perfios, processes more than 12 billion transactions monthly across 80-plus banks in 30-plus countries, giving the platform a data footprint that rivals specialist fintechs built entirely around AI-native architectures. The Mashreq go-live adds a marquee MENA reference customer to that roster at a moment when Gulf sovereign institutions are actively evaluating which AI vendors can operate within their data-residency and Sharia-compliance requirements simultaneously.
Cross-sector read-across: from banking fraud to enterprise AI governance
The convergence angle here extends well beyond financial services. The architectural pattern Clari5 is demonstrating, a generative AI system that is on-premises, auditable, context-grounded, and human-supervised, is precisely what large enterprises in defence procurement, healthcare, and critical infrastructure are demanding before they will sign enterprise AI contracts. Clari5 and Mashreq are, in effect, producing a replicable compliance blueprint at scale.
For the broader AI vendor market, that matters. Pure cloud-based foundation model deployments are running into sovereign data requirements across the GCC, South Asia, and the EU. Vendors who can demonstrate an on-premises or private-cloud governance model with a live banking reference are gaining a structural advantage in procurement conversations outside banking. The Mashreq deployment is a proof point that will be cited in those negotiations.
Capital allocation is also shifting accordingly. Perfios, which acquired Clari5 and operates across 18 countries, represents a category of Indian-origin B2B SaaS platforms that have quietly built compliance-grade infrastructure at a scale that many Western fintechs have not yet matched. As Gulf sovereign wealth funds and regional development finance institutions expand their technology procurement beyond US hyperscalers, platforms with deep MENA and South Asia banking penetration are becoming structurally more attractive investment targets.
The immediate next step for Mashreq will be extending the AI capabilities across additional fraud typologies and, potentially, into anti-money laundering workflows, an adjacent problem where explainability requirements are, if anything, even more stringent. For the wider market, the more consequential question is whether the on-premises, governance-first deployment model Clari5 is advocating becomes the default template for enterprise AI in regulated industries, or whether regulatory harmonisation eventually creates space for cloud-native models to compete on equal terms.