Community Trust ScoreVerified
Central bankers are rattled. Nikhil Rathi, CEO of the UK’s financial watchdog, is pushing hard for new tools and a genuinely collaborative approach to deal with what he sees as a fast-moving threat: agentic artificial intelligence embedding itself into the financial system faster than regulators can track it.
The concern isn’t abstract. AI systems in finance can now operate autonomously — making decisions, executing trades, managing risk — without a human in the loop. That autonomy is the core problem. When something goes wrong, who’s accountable? Regulators built their frameworks around human decision-makers. Those frameworks are struggling badly right now, and Rathi seems pretty clear-eyed about that gap.
Why Agentic AI Scares Regulators
Traditional oversight was built for a slower world. A compliance officer reviews a process. An auditor checks the books. A regulator examines a firm’s risk model. But agentic AI doesn’t wait for a quarterly review cycle. It acts. It adapts. And it can interact with other AI systems across institutions in ways that create knock-on effects nobody planned for. That’s the systemic risk angle that’s keeping central bankers up at night.
The complexity of these systems adds another layer. They’re not transparent by default. You can’t always open the hood and see why a decision was made. For financial regulators whose entire job is maintaining market stability and protecting consumers, that’s a fundamental problem — not a minor inconvenience.
And it’s not just a UK issue. Financial markets are global. AI developers are global. A risk that originates in one jurisdiction can ripple across borders in seconds. Rathi and his counterparts elsewhere know that uncoordinated national responses won’t cut it.
What Regulators Are Actually Considering
Several directions are on the table. AI-specific regulations are one path — rules written not for the financial products AI manages, but for the AI systems themselves. Enhanced transparency requirements are another, forcing firms to make their AI systems more auditable so regulators can actually examine what’s happening inside them.
International collaboration is also a priority. The logic is straightforward: if AI and capital markets are both borderless, the regulatory response probably needs to be too. Getting different jurisdictions to agree on harmonized standards is slow, painful work. But the alternative — a patchwork of national rules that AI systems can route around — seems worse.
There’s also a push to build new monitoring tools. Not just rules on paper, but actual technical mechanisms that let regulators track AI activity in real time, flag anomalies, and verify compliance. That’s a significant build. Unclear yet how far along any of this is in practice.
Rathi’s broader call is for regulators to stop sitting at a distance and start engaging directly with AI developers and financial institutions. That’s a shift. Old-school regulatory culture often meant issuing rules from above without much back-and-forth with industry. The argument now is that you can’t write sensible rules for technology you don’t deeply understand — so you’d better get in the room.
The Innovation-Safety Tightrope
Nobody wants to kill the technology. That’s worth saying clearly. AI does bring real benefits to financial services — faster processing, better fraud detection, more efficient credit decisions. The goal isn’t to shut it down. It’s to make sure the risks don’t outrun the guardrails.
But that balance is genuinely hard to strike. Regulate too loosely and you get systemic fragility. Regulate too tightly and you push innovation offshore or underground. Regulators are aware of both failure modes, and they’re trying to thread a needle that keeps moving.
Getting diverse stakeholders — banks, fintech firms, AI developers, consumer groups, international bodies — to agree on anything is its own challenge. Each group brings different priorities. Banks worry about compliance costs. AI developers worry about regulatory overreach. Consumer advocates want protection. International bodies want coordination. Making those interests converge into workable policy takes time nobody feels they have.
The absence of a unified global standard is still the biggest structural gap. As AI systems grow more capable, the pressure for some kind of international alignment will only increase. Cross-border risks don’t respect the jurisdictional boundaries that regulators are stuck working within.
Rathi’s position is that new tools and direct collaboration aren’t optional extras — they’re the minimum requirement for keeping pace. Whether the regulatory community moves fast enough is a different question. The technology isn’t waiting.
Agentic AI in finance is probably the hardest oversight problem central bankers have faced in a generation. And right now, the frameworks designed to manage it don’t fully exist yet.
Frequently Asked Questions
Who is Nikhil Rathi and what is his role in AI regulation?
Nikhil Rathi is the CEO of the UK’s financial watchdog. He has called for new regulatory tools and a collaborative approach to manage the risks posed by agentic AI in financial services.
What makes agentic AI particularly risky for financial markets?
Agentic AI can operate autonomously without human intervention, raising serious questions about accountability and control — core concerns for regulators tasked with maintaining market stability and protecting consumers.
