The most consequential regulatory document on AI in finance this year was written in London, and its subtext points east. The Mills Review, the FCA’s 147-page study of how AI could reshape retail financial services by 2030, was published in July 2026 as the first review of its kind initiated by a regulator anywhere.

Its seven recommendations are sensible, sequenced and, in several cases, descriptions of work APAC regulators have already done.

The review recommends the FCA enable the foundations for agentic finance, build an AI-enabled supervisory model and scale up its AI Lab. Singapore and Hong Kong didn’t wait for a review to tell them to do any of this.

They’re shipping frameworks, sandboxes and subsidised compute now, and the Mills Review’s own international annex says as much, naming the UK and Singapore as the two most active jurisdictions in building regulatory support for AI adoption in finance.

UK consumers already trust chatbots more than the rules protecting them

The review’s central claim is that retail finance is shifting from human-led, episodic activity to services that are AI-enabled, continuous and delegated.

It maps that shift on an autonomy spectrum in which the human moves from operator to observer, setting boundaries and monitoring outcomes while AI agents act within agreed limits.

The consumer evidence behind it is striking. A survey of more than 5,000 UK consumers conducted for the review found that 20% would be likely to use AI capable of acting autonomously within pre-set goals.

The review reports that around 26% trust general-purpose tools such as ChatGPT, Claude or Gemini for financial advice, despite limited awareness that formal routes to redress don’t apply.

The review’s answer isn’t new rulemaking. It finds the existing outcomes-based framework, anchored by the Consumer Duty and the Senior Managers Regime, sound.

What it asks for instead is plumbing: a perimeter review within three to six months, system-wide oversight of shared models and infrastructure, an Agentic Supervisory Model giving the FCA AI-enabled monitoring tools, and the identity, authorisation and payments foundations that agentic finance requires.

Singapore has already shipped ahead

On 3 July 2026, MAS and industry partners released Safeguards for Agentic Finance at Runtime, or SAFR, a framework specifying how AI agent actions in finance can be authorised, validated in real time, escalated to human review and recorded before execution.

Industry members have already applied it to agent-assisted payments, treasury operations and wealth advisory workflows.

SAFR didn’t arrive in isolation. In March, MAS concluded phase two of Project MindForge with an AI risk management toolkit built with a consortium of 24 banks, insurers and capital markets firms, covering traditional, generative and agentic AI.

Kenneth Gay, chief FinTech officer at MAS, said the toolkit marked “a major step forward in our journey to ensure the responsible adoption of AI” in finance. Singapore’s regulator is describing a build already in progress, co-developed with the institutions that will run on it.

Hong Kong is turning policy intent into operational infrastructure

Hong Kong offers the second proof point. On 5 March 2026, the HKMA, SFC, Insurance Authority and MPFA jointly launched the GenA.I. Sandbox++, expanding the 2024 banking-only sandbox across securities, asset and wealth management, insurance, pensions and stored value facilities.

Participants get access to GPU computing at Cyberport’s AI Supercomputing Centre alongside supervisory guidance, with the programme targeting risk management, anti-fraud and customer experience use cases.

The Mills Review recommends scaling the FCA’s AI Lab. Hong Kong has bolted its sandbox to state-backed compute and run it across every regulated sector at once.

Four regulators co-signing one experimentation framework is the system-wide coordination the review calls for, delivered as an application form with a 30 June 2026 deadline rather than a recommendation.

The scenario London is modelling is already live in Asia

The review’s annex on agentic infrastructure concedes that the consumer behaviour it forecasts for the UK is already operating at scale in Asia.

It cites India’s UPI integrating with AI assistants, and notes Alipay’s agentic payment infrastructure serving 100 million users and processing 120 million agent-initiated transactions in a single week in February 2026. Those are the volumes London is modelling as a 2030 scenario.

There’s a competition warning in the review that APAC readers should take more seriously than anyone. It argues that control of the AI-mediated customer interface, the assistant or agent through which consumers search, compare and transact, could become a major source of market power, deciding which products are visible and where value is captured.

In a region where super-apps already own the customer relationship, and Alipay is running agent-initiated payments at nine-figure weekly volumes, this is a live description of how distribution is consolidating, and regional regulators have said far less about it than the review does.

The Mills Review’s analysis of where risk migrates as autonomy grows, from harm within a single firm to correlated, system-wide failure across shared models and infrastructure, is the sharpest regulatory treatment of the subject yet published, and its Agentic Supervisory Model goes further than anything APAC regulators have committed to.

Singapore’s SAFR governs how firms deploy agents; nobody in the region has yet built the AI-enabled supervision of the whole system that the review says will be necessary.

The review recommends the FCA examine, within three to six months, how general-purpose AI tools shaping financial decisions outside the regulated perimeter should be treated. If the FCA’s perimeter review produces a workable answer, it’ll be the first, and APAC regulators will be reading it as closely as UK firms.

The likelier outcome is convergence. The review recommends the FCA work internationally on agentic standards, and the obvious counterparts are the jurisdictions it benchmarks.

For APAC institutions, the practical reading is this: the UK has just published the clearest map of where AI-enabled finance is heading, and the regulators closest to home are already laying the road.

Firms that treat SAFR, MindForge and the Sandbox++ as early drafts of a global baseline will be better placed than those waiting for their regulator’s version of the Review.