Nikhil Rathi, chief executive of Britain's Financial Conduct Authority (FCA), said Thursday that AI agents working with growing autonomy will force his supervisors to decide "when agentic coordination becomes market manipulation," according to the speech text the regulator published. The remark tells banks, asset managers and trading firms in London that the FCA expects to rule case by case on how autonomous software behaves in markets, before any law spells it out.
What Happened
Rathi delivered "Market Integrity in a Digital Age" at the City of London's annual City Dinner at Mansion House. The FCA notes the posted text is the speech as drafted and may differ from delivery.
He said threats to market integrity are as serious as he has seen them, listing unprecedented volatility, concentration risk, new forms of market abuse and cross-border financial crime, all intensified by fast-emerging technology. Frontier AI models, the most capable general-purpose systems, are making it easier for AI agents to carry out tasks increasingly autonomously, he said, calling British firms fast adopters. He expects personal AI agents, such as Meta's Muse app, to reach U.K. consumers soon.
The FCA is using the same tools. Rathi said agents can assess cases, synthesize regulatory returns and intelligence, and comb millions of rows of data for missing or concerning information, freeing staff for the hardest calls, including the manipulation question and judging when a risk becomes system-wide.
On cybersecurity, he cited the U.K.'s AI minister as saying it is likely that British systems will be breached at some point. He stressed reporting and learning from near misses, insurance across firms, customers and supply chains, and investment in recovery, because prevention will not always work.
Why It Matters
Market-abuse rules were written around human traders and human intent. When separate agents react to the same signals at machine speed, their combined behavior can look coordinated without anyone agreeing to it. Rathi did not say where the line falls, which leaves firms that deploy trading or customer-facing agents in U.K. markets, including U.S. firms with London operations, exposed to supervisory judgment rather than a published test.
He also said the FCA will write rules where needed, including aligning with national or global AI standards, which the prime minister has said will be a focus of the U.K.'s coming G20 presidency. He urged more assertive use of existing powers, such as the regime for critical third parties, the outside technology providers the financial system depends on.
Context
The speech follows remarks Monday by Jessica Rusu, the FCA's chief data, information and intelligence officer, who said firms should report significant AI failures and near misses, The Primary Wire reported.
What to Watch
Whether the FCA issues guidance on agent-driven trading, what AI standards emerge under the G20 presidency and the joint FCA and Bank of England roadmap on digital assets Rathi described.
