Marsh & McLennan is pressing US regulators to treat credit insurance on AI loans the way banks want: as genuine capital relief. This matters to UK regulated firms because the FCA and PRA will face identical pressure within months, and they are unprepared. Right now, UK banks hold full capital against loans to AI infrastructure firms even when those loans are insured—because regulators treat the insurance as risk transfer in theory but not in capital practice. If Marsh wins in America, the argument becomes irresistible here. And if the FCA gives way without embedding proper AI-specific risk controls into its capital framework, UK lenders will flood capital into AI lending with minimal skin in the game. That is not prudent regulation.

This story is part of a larger pattern: regulators are reactive, not anticipatory. The AI infrastructure lending market has moved faster than regulatory tooling can follow. Banks are creating portfolios of loans to companies that may not exist in three years. The collateral is often intellectual property that has no secondary market. The borrowers' revenue models depend on regulatory decisions in multiple jurisdictions. Yet the insurance industry—and now Marsh—is packaging these loans as 'credit risk' that can be hedged like any other debt exposure. It cannot. AI lending is not credit risk in the traditional sense. It is model risk, regulatory risk, and technology obsolescence risk wearing a credit risk mask. The FCA's PRA SS1/23 guidance on outsourcing and third-party AI risk does not even mention lending. That is the real gap.

Trovix's view: this story reveals why generic AI tools fail at regulatory compliance. Products like Harvey or Copilot are excellent at legal research and document drafting, but they miss the structural risk entirely. They cannot flag that a loan portfolio to AI companies is fundamentally different from a loan portfolio to telecoms operators. They cannot integrate regulatory capital rules with AI risk assessment because they do not know which questions to ask. The honest answer is that credit insurance on AI loans should carry higher capital charges than traditional credit insurance, not lower ones. Regulators should require banks and insurers to use AI risk frameworks—tools that can monitor model drift, regulatory change, and technology obsolescence in the borrower's stack—before any capital relief is granted. Trovix Watch exists precisely because regulators move in lockstep and firms need to see the pattern before the rule lands.

If you are a mid-market bank, insurer, or finance house: do not assume capital rules will shift in your favour. Prepare now for the opposite. Map every AI-related loan or investment in your portfolio. Identify which ones depend on regulatory decisions you cannot control. Build an audit trail showing you understood the AI-specific risks—not just the credit risk. When the FCA eventually clarifies its position (and it will), firms that can prove they assessed AI lending as a distinct category will have regulatory credibility. Firms that treated it as standard commercial lending will face capital surcharges. The PRA's approach to model risk governance in banks, and the ICO's emerging frameworks on algorithmic accountability, are your real regulatory north star here, not Marsh's lobbying.

Source: Bloomberg News