Central banks move digital money and AI from research to rulebooks

Hong Kong: accountability first

The Hong Kong Monetary Authority moved AI in banking from sandbox to supervision on October 15 with its 'ART' Framework — Accountability, Resiliency, Trust — a package of new supervisory guidance for responsible AI adoption. The framework comes with operational plumbing: an AI training data-sharing platform being built with Cyberport and the University of Hong Kong, expanded testing for agentic AI through the GenA.I. Sandbox++, and a planned discussion paper on agentic AI governance. The sequencing matters — guidance first, then shared data infrastructure, then deeper testing of the riskiest use cases. m:…7301

Arthur Yuen said, "The progression from generative to agentic A.I. represents an important new stage of innovation." and the Sandbox numbers back the urgency: over 60% of use cases in the second GenA.I. Sandbox cohort focused on agentic AI. Yuen added, "By pairing innovation with clear accountability, resilient operating models and effective human oversight, banks can fully realise the benefits of A.I." The ART framing is deliberate — the HKMA is telling banks that the license to deploy agentic systems will be conditioned on demonstrable accountability, not just model performance. m:…7301

Frankfurt: the limits that backfire

The European Central Bank's contribution, also dated October 15, is research with a supervisory edge. An ECB working paper on the behavioral effects of CBDC holding limits finds that the limits themselves can change behavior: high limits can trigger an 'attraction effect' that inadvertently increases user balances — the guardrail drawing people toward the thing it caps. The study, run across four euro-area countries, is a caution for any central bank designing a digital euro with holding caps as the financial-stability backstop. m:…6916

The paper's proposed fix is mechanical rather than moral: automatic 'reverse waterfall' conversion — sweeping excess balances into linked accounts — can boost adoption while keeping individual holdings within desired bounds. For digital-money teams, the takeaway is that CBDC design is now an exercise in behavioral engineering as much as monetary plumbing, and the ECB is publishing its lab notes. m:…6916

The global template

The third piece is the global template. The Bank of Japan published the Financial Stability Board's consultation report, 'Regulatory and Supervisory Modernisation Principles,' which sets out eight high-level principles to help financial authorities update their regulatory frameworks while balancing sustainable growth with financial stability. The FSB is seeking public feedback — the Bank of Japan is facilitating the collection of comments, with the consultation open until January 8, 2027 — and the report leans on case studies so member authorities can learn from each other's modernization efforts. The FSB is seeking to harmonize regulatory and supervisory frameworks so they remain forward-looking and adaptive to material risks, and member authorities are encouraged to use the case studies for mutual learning across jurisdictions. m:…5913

The pattern across all three is a shift from permission to prescription. Hong Kong is writing the operating rules for agentic AI in banks; Frankfurt is publishing the behavioral evidence that will shape CBDC caps; the FSB is supplying the common language for modernizing rulebooks without losing the stability plot. The era of central banks studying digital money and AI from a distance is ending — the rulebooks are being drafted in public, and the comment windows are open. m:…5913 m:…7301