The head of Britain's financial regulator used the City of London's biggest annual dinner to deliver an FCA AI warning. Speaking at Mansion House on Thursday, 22 October, Financial Conduct Authority chief executive Nikhil Rathi said the old problems of finance are being sped up by new technology, and that regulators built for a slower world will fall behind unless they change how they work.
In the drafted text of the speech, published by the FCA under the title "Market integrity in a digital age", Rathi named four threats: "unprecedented volatility, concentration risks, new forms of market abuse and rampant cross-border financial crime". He then added the line that will be quoted most: "fast-emerging technologies putting these on steroids."
The FCA notes that the published text is the speech as drafted and may differ from what was delivered on the night.
What the FCA AI warning covers
Rathi did not announce a new rulebook for artificial intelligence. The speech is a statement of where the regulator thinks the risk is moving and how it plans to respond. Four areas stood out.
Speed and volatility
Markets now move at speeds and volumes that supervisors were not designed to watch in real time. Tech Times, which reported on the speech, said Rathi pointed to AI-driven trading that carries on over weekends, when traditional market oversight is thin.
Concentration on a few providers
Banks, brokers and payment firms increasingly run on the same small group of cloud and AI suppliers. When one fails, the effect spreads. Tech Times reported that Rathi cited a Google Cloud outage in the UK that disrupted financial services, including the app-based bank Revolut. Under the Financial Services and Markets Act 2023, the FCA and the Prudential Regulation Authority can already oversee "critical third parties" whose failure would threaten stability.
New forms of market abuse
The speech flagged "insider trading on offshore gambling or prediction markets", activity that largely sits outside the FCA's formal perimeter. Rathi also raised a harder question for supervisors: deciding "when agentic coordination becomes market manipulation". If several AI trading agents learn to act in step, a regulator has to judge whether that is collusion, coincidence or simply efficient pricing. The FCA has not yet published guidance on how the UK Market Abuse Regulation applies to agentic trading.
Crime that crosses borders
Fraud and money laundering networks move money through several countries in minutes. "No single agency can rise to today's challenges alone," Rathi said, pointing to recent action by US authorities on anti-money laundering failures and to joint work with international counterparts.
Meta Muse is the test case
The most specific part of the speech dealt with personal AI agents. "Take Meta Muse, launched 6 weeks ago, one of the fastest-downloaded apps in US history," Rathi said. "Personal agents are profoundly changing how consumers interact with financial and health services." He added that "such applications will reach UK consumers soon".
The FCA is working on Muse through the Digital Regulation Cooperation Forum, which brings together the FCA, the Information Commissioner's Office, Ofcom and the Competition and Markets Authority. That matters because an agent that books, buys and pays on a user's behalf touches data protection, competition and consumer protection at the same time. The ICO has already opened its own inquiry into AI agents, as we reported in our piece on the ICO's call for evidence on agentic AI.
The regulator is using agents too
Rathi described the FCA as an "agentic regulator" in the making. According to the speech, AI tools give supervisors "a richer risk picture", and agents "can assess cases, synthesise returns and intelligence" and "parse millions of rows of data and identify missing or concerning information".
Tech Times reported two concrete uses: automatically checking firms' regulatory returns against market data, and testing an AI agent as a "first responder" in wholesale market surveillance. Human supervisors keep the final say on enforcement.
The FCA also published guidance on good and poor practice in firms' use of AI on 19 October, according to Tech Times.
What comes next
- Digital assets: Rathi said an FCA and Bank of England roadmap on digital assets is coming "soon".
- Identity checks: Johan Sekora, the FCA's new chief operating officer, will lead work to "look beyond know your customer (KYC)" processes.
- Standards: the regulator wants to align with national and global AI standards, which the speech linked to the UK's upcoming G20 presidency.
- Compliance load: deputy chief executive Sarah Pritchard will work with industry to cut duplicated compliance processes, Tech Times reported.
Why it matters beyond London
The FCA supervises tens of thousands of firms, and its approach is closely watched by regulators in Africa, the Gulf and Asia that model their rules on British practice. Two points from the speech travel well.
First, the FCA is treating AI as a force multiplier for existing risks rather than as a separate category that needs its own law. That keeps enforcement inside current rules on market abuse, outsourcing and consumer duty.
Second, it is betting that the answer to faster crime is faster supervision, using the same kind of agents that criminals and traders are adopting. That raises its own questions about accuracy, accountability and appeals when a machine flags a firm. Rathi's speech did not answer them, and the agentic market manipulation guidance he hinted at will be the first real test of how far the regulator is willing to go.
For firms building or deploying AI agents that move money, the practical message is simple: the FCA expects them to show how those agents are controlled, logged and stopped, before a supervisor's own agent comes asking.
