Share
The UK’s Financial Conduct Authority published its most significant AI policy document to date on July 6, 2026. The Mills Review, led by FCA Executive Director Sheldon Mills, examined how AI will reshape retail financial services by 2030. Its central conclusion is that the transition from AI as a decision-support tool to AI as an autonomous decision-maker is already underway.
The Mills Review explores how AI could reshape retail financial services for consumers, firms, markets and regulators.
The findings and recommendations from the independent report build on the FCA's strong foundations and how we can harness AI to deliver better outcomes for…
— Financial Conduct Authority (@TheFCA) July 6, 2026
FCA Chief Executive Nikhil Rathi described the next phase clearly, saying it will be made up of systems that do not just support financial decisions but coordinate and transact on behalf of consumers without human authorization at each step.
The Mills Review sets out seven recommendations. We highlight two of those for their relevance to crypto infrastructure.
FCA’s first crypto-related recommendation addresses the regulatory perimeter. If consumers increasingly depend on AI systems from a small number of large technology companies to make financial decisions, those companies may exert influence comparable to regulated firms without carrying equivalent obligations. The FCA noted that its Critical Third Parties regime is technology-agnostic and could capture major AI and cloud providers where designation criteria are met.
The second recommendation dwells on the agentic finance infrastructure question. The Review asks who will control the primary customer relationship by 2030, incumbent financial services firms, Big Tech, specialist AI intermediaries, or consumers’ own AI agents. The FCA’s implicit position is that it does not yet know and is actively seeking input from the industry.
As a regulator, the FCA is also building its own agentic capabilities. It is deploying agentic AI as a first responder for wholesale market surveillance, processing a billion rows of data per day to detect market abuse faster than human supervisors. It has launched an Agentic Academy alongside its AI Lab and Supercharged Sandbox, where firms can test solutions using real-world data with partners including Nvidia and Google.
The infrastructure the FCA is asking about already exists, and has been building up for over two years.
AI activity on XRP Ledger now has a home.
Today, we’re launching the XRPL AI Hub, featuring our ecosystem launch partner @virtuals_io.
A single destination for agents, AI projects, tools, and payment services building on the XRPL.https://t.co/gwLZblD0wf pic.twitter.com/QXewkkYlnR
— t54.ai (@t54ai) July 8, 2026
Ripple launched its XRPL AI Starter Kit in June 2026, giving developers tools for agent payments using XRP and RLUSD. Solana hosts 65% of AI agent payment volume according to Messari’s Q1 2026 data, with sub-cent fees and 400ms finality making it the most economically viable chain for high-frequency agent transactions.
Coinbase’s x402 protocol enables agents to make direct payments over the internet using stablecoins. Google’s interest in x402, which Hoskinson referenced at Consensus Miami, signals that the largest consumer AI platform is already evaluating crypto payment rails for agent commerce.
The intersection of agentic AI and tokenized assets creates a new regulatory gap: an AI agent capable of holding and spending tokenized assets in real time on behalf of a consumer across a programmable blockchain.
Until now, no regulatory framework was designed for such a scenario. The Mills Review’s seven recommendations represent the FCA’s first attempt to identify what needs to change.
Share
preferred on
