Sri Lanka is closer to an AI-ready economy than it realises, but central banks that adopt the technology without governance risk the public trust they exist to protect, the International Monetary Fund’s chief information officer said in Colombo on Friday.
Shirin Hamid, CIO and Director of the IMF’s Information Technology Department, was delivering the Central Bank of Sri Lanka’s 76th Anniversary Oration, titled “The Intelligent Central Bank: Building Trust in the Age of AI.”
“The intelligent central bank is not the bank with the most AI. It is the bank whose AI can be trusted, and can ensure that any AI that touches the financial system is explainable, governed, secure, ethical, and human-supervised.”
The case for optimism
Hamid said Sri Lanka had used its recovery “not just to stabilize, but also to build,” pointing to the country’s digital economy blueprint. NewsFirst reported the figures she cited: interbank electronic fund transfers now make up nearly 70 percent of retail transaction volumes, up from single digits a few years ago, and more than 60 percent of government revenue is collected digitally. Digital literacy stands at 72 percent. A national digital identity system and a national data exchange platform are both in progress.
She argued that being a smaller and later mover is an advantage: “You’re not dragging 50 years of legacy systems behind you.” Leapfrogging, she said, “is not a slogan here. It is an option, and it is yours to exercise, but with judgment.”
The case for caution
The warnings were specific. EconomyNext reported that Hamid identified three threats to financial institutions: impersonation using cloned audio and video, autonomous AI “swarms” attacking at machine speed, and indirect breaches that manipulate AI assistants already operating inside bank firewalls. As little as three seconds of recorded voice is now enough to produce an indistinguishable clone.
On data, she disclosed that when the IMF tested commercial generative AI tools on economic figures, systems such as ChatGPT returned inaccurate GDP numbers up to two-thirds of the time. Because plausibly wrong numbers damage institutional credibility, the Fund built its own system, StatGPT, which queries verified data directly rather than letting a model guess.
Four pillars
Hamid urged central banks to ground the transition in dependable data, trustworthy systems, operational resilience and adaptive governance — and never to treat AI as an IT department project. Human judgment, she said, remains the scarcest resource.
“For every AI-influenced decision, we need a named human who is answerable. We should aim to augment judgment first, and grant autonomy only where it has been earned.”
The mission of a central bank, she concluded, is unchanged: safeguarding public confidence “from the vault to the algorithm.”
