Dubai regulator chief says finance AI gains hinge on watchdogs

Artificial intelligence could speed up financial compliance, but Matthew White, chief executive of Dubai’s Virtual Assets Regulatory Authority, says regulators may determine how much of that efficiency reaches the industry. In a Fortune commentary, White argued that financial watchdogs are both users of AI and gatekeepers for its adoption by banks, crypto firms and other licensed businesses.

White leads VARA, which he described as the world’s first regulator dedicated to virtual assets. He said his experience building the authority, after more than a decade advising financial institutions on cybersecurity and technology risk at PwC, showed that supervisory infrastructure can matter as much as rule-writing.

White said large banks have spent more than 10 years and billions of dollars on regulatory technology, with uneven results. He cited Thomson Reuters’ annual Cost of Compliance surveys as showing compliance budgets and staffing rising steadily while productivity gains remained limited.

Large language models triggered broad experimentation in 2023, White said, but he argued that compliance-critical uses have seen little production deployment because mistakes in regulated processes carry high costs. Among public authorities, he said the gap is wider: Singapore’s Monetary Authority, the U.K. Financial Conduct Authority, the Hong Kong Monetary Authority and the Bank of England run serious supervisory-technology programs, while many other regulators still rely heavily on PDF rulebooks, onsite sampling and email-based reviews.

Virtual assets push a different model

White said virtual assets are forcing regulators to rethink supervision because the systems are programmable, cross-border and continuously active. In his view, quarterly inspections and paperwork-based reviews do not fit smart contracts or autonomous lending protocols.

He said the sector requires on-chain audits, live monitoring and programmable compliance. White argued that tools first built for crypto oversight could become part of mainstream finance as tokenization spreads to government bonds, money market funds, equities, real estate and private credit.

White cited McKinsey’s 2024 estimate that tokenized financial assets could reach $2 trillion by 2030. He also pointed to the Bank for International Settlements’ Project Agorá, which he said is in testing with seven central banks and more than 40 financial institutions on tokenized wholesale settlement.

Two timelines for AI supervision

White laid out two expected phases for AI in regulation. Over the next 12 to 24 months, he said authorization reviews that now take weeks could be shortened to days, and AI could reduce compliance alerts by reading transactions in context rather than applying fixed rules.

He cited industry research showing false-positive rates in compliance alerts between 90% and 95%. Over a three-to-five-year period, White said regulators and licensees could shift from periodic filings to continuous data exchange, weakening the case for sample-based inspections.

That shift, White said, would put pressure on bank compliance teams, especially junior analyst roles and older compliance software. He framed the central constraint as regulatory readiness rather than corporate demand.

White also said legal accountability remains unresolved. Compliance law still assumes human decision-making, he argued, raising questions about responsibility when AI helps produce a suspicious activity report or contributes to a fitness-and-propriety assessment.

He said Europe’s MiCA implementation and the United States’ CLARITY Act framework are among the policy moments that could answer those questions. White warned that jurisdictions that keep requiring PDF returns and human-mediated supervisory cycles could lose business to regulators that support real-time data and AI-enabled oversight.

This story draws on original reporting from Fortune.