Member of the Treasury Committee, House of Commons. Former City Minister from 2018 to 2022, leading the reforms later enacted through the Financial Services and Markets Act 2023.
“The existing framework was built for a world in which senior managers make decisions and can be held directly accountable for them, not for a world in which autonomous agents act continuously, and in which a small number of cloud and model providers sit beneath the entire sector. The Mills Review, the Bank's Financial Stability Report and Nikhil Rathi's speech in June all ask, in different ways, whether technology-neutrality is sufficient when the technology itself changes the nature of accountability and concentration risk. That deserves a serious answer.”
— Rt Hon John Glen MP
Accountability requirements move out of statute and into the regulators' rulebooks, so future changes to how AI is supervised will not have to wait for Parliament.
Sets out the questions supervisors are being equipped to ask, which makes it the closest thing yet to a preview of your next AI conversation with the FCA.
Establishes materiality and risk as separate tests, so a small use case with real autonomy stops being a small governance problem.
The FCA signals it will reach for market-wide powers routinely rather than exceptionally, moving intervention beyond your own supervisory relationship.
The Bank asks in public whether technology-agnostic rules remain sufficient, the clearest signal so far that AI-specific requirements are on the table.
Recommends the regulator supervise with AI as firms adopt it, replacing periodic document review with continuous, data-driven oversight.
Treats AI as a source of system-wide financial risk rather than a firm-level operational one, which changes who is watching and what they are watching for.
Brings the shared layer beneath the market under direct supervision, while leaving responsibility for depending on it firmly with your firm.
Proposes sector-wide incident sharing and third-party assurance, pointing towards supervision built on pooled evidence rather than firm-by-firm returns.
Names structured governance over AI as good practice and concentrated, thinly evidenced ownership as poor, giving you the benchmark before the guidance lands.
Shows adoption still limited but accelerating, which is why supervisory expectations are being set now rather than after the fact.
Finds vulnerability discovery outpacing remediation, making the governance around the model, not the model itself, the thing under examination.
Recent developments point towards more continuous, system-wide supervision as AI becomes more autonomous. Regulators are considering how supervision should adapt as AI operates across increasingly interconnected parts of the financial system.
Greater autonomy makes senior-manager accountability and meaningful human oversight harder to exercise and to evidence. Probabilistic systems make point-in-time assessment less reliable, and shared reliance on common cloud, model and data providers creates concentration and resilience risk across firms.
The Financial Services and Markets Bill would remove statutory requirements from the SM&CR and give the FCA and PRA more flexibility over how the accountability framework operates. Not an AI reform, but it gives regulators room to adapt one.
General-purpose AI tools can increasingly influence savings, investment and pension decisions while potentially sitting outside the FCA's existing perimeter. The Mills Review recommends examining whether advice-like outputs from general-purpose AI create a regulatory gap.
None of this yet amounts to a separate rulebook for AI. The UK's principles-based, outcomes-focused framework remains the foundation. What is being examined is which parts of that framework, and the way it is supervised, need to evolve.
The FCA's AI good and poor practice, expected later in 2026. The progress of the Financial Services and Markets Bill, particularly SM&CR reform. The next Bank of England and FCA adoption survey. The FSB's final Sound Practices.
Head of UK Compliance
Managing Director and Head of EMEA Compliance & Financial Crime
MLRO
Head of Legal
Regulatory AI Architect
CEO and Co-Founder
Partner
Partner and Head of Wealth Management Regulatory
Adapting supervision to AI may require more targeted rules that depart from the technology-neutral principle the UK framework has relied on.
Senior-manager accountability, human oversight and operational resilience were all built around people making individual decisions.
The Mills Review's agentic supervisory model and the FCA's readiness to use system-wide powers point the same way.
Firms will need new ways to evidence that oversight is meaningful and ongoing, rather than a moment of approval.
