UK banking is sleep walking into the greatest structural disruption in its history, and it’s funding the disruption itself.
The US Government’s suspension of Anthropic’s Mythos and Fable models on national security grounds should serve as a wake-up call for the UK and for financial services in particular, as it races to capitalise on AI. The US has flagged national security concerns with this particular model. But what this does is expose the very clear safety and security issues around the mass adoption of a generalist AI model built entirely in another geography to be utilised in a highly regulated sector in the UK. This is simply not a rigorous enough approach for the banking sector.
Banks, insurers and wealth managers are racing to adopt artificial intelligence. Boards are approving AI contracts. Executives are urged to automate, modernise and show credible AI strategies. Individually, these decisions look sensible. Collectively, they may be dangerous.
The industry thinks it is buying productivity. It may also be financing its own disintermediation.
What is being built is not simply better software. It is a new distribution layer for financial services, one that sits above institutions themselves. Once that layer consolidates, the balance of power changes.
The shift from institution to interface
For decades, financial institutions owned the customer relationship. They controlled the interface through which consumers understood and managed current accounts, mortgages, savings, pensions, investments, insurance and advice. The products mattered, but so did the trusted channel.
AI threatens to break that model. Not because consumers will stop needing regulated institutions, but because they may stop interacting with them directly.
The first stage is already visible. Consumers increasingly prefer tools that simplify fragmented financial lives. An AI assistant that can understand spending, borrowing, pensions, investments and life goals at once is more useful than multiple banking apps. Consumers will adopt these systems because the experience is better.
At that point, power begins to shift. Once an AI Agent becomes the primary interface for a consumer’s financial life, the institution behind the product becomes less important than the system controlling discovery, recommendation and decision making. The bank, insurer or wealth manager retains the balance sheet exposure, operational complexity and regulatory liability. The platform controlling customer intent captures the economics of distribution.
The Uberisation of financial services
The analogy is uncomfortable: this is the Uberisation of financial services. Uber did not need to own cars to control the economics of transport. They won by owning the interface, shaping consumer choice and reducing suppliers to participants inside a platform controlled market.
The same logic could apply to finance. A bank may still provide the mortgage, savings account or investment product. It may still hold the capital, meet regulatory obligations and absorb operational risk. But if the customer asks an AI assistant what to do, and that assistant compares, recommends, switches and executes, then the value has moved. The product provider remains regulated infrastructure. The platform owns demand.
Open banking was meant to change the whole distribution layer. That proved to be a myth, but the dial has moved on considerably since then.
The economic exposure is substantial. The UK does not have to lose its financial services sector for this to matter. Losing the distribution layer alone, the economics attached to advice, recommendation, switching, customer data and cross-sell, could plausibly put £20bn-£30bn of annual value at risk. In a severe case, where hyperscaler controlled assistants become the dominant financial interface, the exposure could rise towards £40bn-£50bn, including over half a million jobs in the industry.
That is why this is not simply a technology procurement issue. It is a national economic resilience issue.
Both London and Edinburgh are leading financial centres, employers and AI hubs. That should be a national advantage, if the value created by applying AI to finance is retained within the UK-regulated financial system, rather than migrating to global platforms.
Regulation could inadvertently accelerate the shift. The FCA’s work on the advice-guidance boundary is intended to improve access to support for consumers. But if simplified guidance becomes AI mediated at hyperscaler scale before the UK has clear guardrails around governance, accountability and data use, reform may open the door to a major transfer of economic power.
This is not an argument against AI. AI can make advice more accessible, operations more efficient and fraud detection more effective. The question is not whether AI should be used. The question really is who owns the relationship once it is.
A call to action for UK financial services
First, the industry should build a UK controlled AI trust layer. Second, financial data should be treated as strategic infrastructure, not a oneway supply chain into platforms that face neither the same obligations nor the same national stakes. Third, the regulatory boundary for AI intermediation must be clarified. If a system recommends, ranks, nudges, switches or executes financial decisions, it is part of the distribution chain and should carry appropriate obligations.
Finally, UK firms should collaborate on shared AI infrastructure for fraud prevention, identity, compliance, vulnerability detection and assurance.
The future battleground will not be product manufacturing alone. It will be ownership of the customer relationship. If that relationship migrates elsewhere, the economics follow.
UK finance still has time to shape this outcome. But it should stop pretending every AI procurement decision is just a technology upgrade. Some are building the architecture of the industry’s future. The question is whether that future will be owned by UK regulated financial institutions, or rented back from Big Tech (when we are allowed access to it).
