AI in action · Regulation
Hong Kong's Regulators Turn to AI Agents in Finance
On November 2 the HKMA launched Project SAGE for AI agents in banking, and the SFC found 30% of large brokers already use agentic AI. Supervision is catching up with practice.
On Monday, November 2, the first day of Hong Kong FinTech Week, both of the city's main financial regulators published work on the same subject: AI agents, software that acts on a customer's or a firm's behalf rather than just answering questions. The Hong Kong Monetary Authority, which supervises banks, launched Project SAGE, short for Securing Agentic Finance with Governance and Enablement. The Securities and Futures Commission, which supervises brokers and asset managers, published its first Large Firm AI Readiness Survey.
SAGE has four workstreams, according to the HKMA: testing real use cases, from agents that shop and pay for consumers to agents that run corporate procurement and treasury; verifying an agent's identity and the authority it has been given, which the HKMA calls "Know Your Agent"; interoperability standards so agents from different providers can deal with each other safely; and the legal framework, covering governance, liability and consumer protection. A 22-member expert group supports it, including HSBC, Standard Chartered, Bank of China (Hong Kong), Visa, Mastercard, UnionPay, Google Cloud, Amazon Web Services and Clifford Chance, Fintech News Hong Kong reported. "Agentic finance has the potential to transform how financial services are delivered and utilised," said Eddie Yue, the HKMA's chief executive.
The SFC's survey shows how far ahead of the rulebook the industry already is. It covered 20 large firms with 78 licensed affiliates, which together handled more than 75% of the securities transactions reported by licensed corporations in the first half of 2026. Thirty per cent of the firms already use agentic AI in their operations and another 25% are piloting it, for tasks that include drafting, software development, operational workflows and "some trading-related activities". Ninety-five per cent expect their AI use to grow over the next year, 90% have connected AI tools to internal systems and 90% are using or evaluating frontier AI models for cyber defence.
The SFC's answer is not to require sign-off on every step. Its executive director of intermediaries, Eric Yip, called the approach "safety-first, human-anchored and collaboration-led": firms should set clear objectives and limits on what each agent can access and do, monitor it and be able to intervene, with controls proportionate to the agent's authority and potential impact. Manual approval of every output, the regulator said, is not required. The four priorities it set out are investor protection, cyber resilience, governance and building supervisors' own expertise.
Hong Kong's banking regulator issued its broader AI guidance only on October 16, so the city has gone from principles to agent-specific work in under three weeks. The liability question in SAGE's fourth workstream, who pays when an agent authorised by one bank acts on an instruction from another AI, is one that regulators elsewhere have so far only posed, not settled.
Sources: Hong Kong Monetary Authority — HKMA launches Project SAGE and establishes Expert Group for agentic finance (Nov 2, 2026); Securities and Futures Commission — SFC survey highlights wider AI adoption and challenges to responsible scaling (Nov 2, 2026); Mondo Visione — Hong Kong SFC survey highlights wider AI adoption (Nov 2, 2026); Fintech News Hong Kong — 11 financial institutions join HKMA expert group on agentic finance (Nov 2026)