AI in action · Regulation

When Does "Agentic Coordination" Become Market Manipulation? The FCA's Chief Executive Put the Question to the City at Mansion House

The FCA's Nikhil Rathi told the City on October 22, 2026 that threats to market integrity are "as serious as I have seen them," with AI agents a cause. He spoke at the annual Mansion House City Dinner. Frontier models, he said, make it easier for agents to act increasingly autonomously, and he named the point "when agentic coordination becomes market manipulation" as a risk the regulator must be ready for. He cited Meta's Muse as a sign of what UK consumers will soon use, said FCA supervisors already work with AI agents, and promised to align with any national or global AI standards set under the UK's G20 presidency.

Oct 23, 2026

The annual City Dinner at Mansion House is where the FCA's chief executive tells the financial industry what the regulator is worried about. On October 22, 2026, Nikhil Rathi's answer began with technology. "Vast opportunities," he said in the drafted text the FCA published, "but the threats to market integrity are as serious as I have seen them," listing unprecedented volatility, concentration risks, new forms of market abuse and rampant cross-border financial crime, with fast-emerging technologies "putting these on steroids." His central claim was that the UK's regulatory system moves too slowly for the pace of that change, and that the FCA must focus on outcomes and use its existing powers more assertively.

AI agents ran through the speech. "Frontier AI models are here," Rathi said, "making it easier for AI agents to carry out tasks increasingly autonomously," and UK firms are fast adopters. He pointed to Meta's Muse, a personal agent app launched about six weeks earlier, as an example of how consumers' use of financial and health services is changing, and said he expects such apps to reach UK consumers soon. Among the integrity risks he named, one phrase stood out for markets: the moment "when agentic coordination becomes market manipulation." He did not define it. He did not need to: academic work published this year has shown trading agents learning to coordinate prices without communicating, and existing market-abuse law is built around intent and agreement that a model may not have.

The regulator is using the same tools on itself. Rathi said FCA supervisors can use AI agents to assess cases, summarise regulatory returns and intelligence, and find missing or concerning information in large datasets, freeing staff for the harder judgements. "Expanded capability means we'll find more risks," he added, which is why the FCA will have to be open about which risks justify its resources. He also noted that the UK's AI minister considers a breach of UK systems likely at some point, an argument for near-miss reporting and investment in recovery rather than prevention alone.

The rest of the speech fits around that theme. Rathi flagged insider trading on offshore gambling or prediction markets as a system-wide risk; questioned whether the "three lines of defence" model of compliance still works; said as much as 8% of banks' workforces are dedicated to fighting financial crime and that those resources should deliver better outcomes; and said a new FCA chief operating officer, Johan Sekora, will lead work on looking beyond know-your-customer processes that collect large amounts of data without finding actual money laundering. On standards, the FCA will align with any national or global AI rules, which the Prime Minister has said will be a focus of the UK's G20 presidency.

Read alongside the FCA's other October publications, the shape of British AI policy is now clear. There will be no AI rulebook; there will be examples of good and poor practice, near-miss reporting, AI-assisted supervision and, when a market-abuse case involving agents arrives, enforcement under the rules that already exist. Rathi's question about agentic coordination is the one that framework will be tested on first, because it is the case where existing law fits least well.

Sources: FCA — Market integrity in a digital age, speech by Nikhil Rathi at the City Dinner, Mansion House (Oct 22, 2026); FCA — Building the future of finance in the age of AI, speech (Oct 19, 2026)