FCA's Landmark AI Review Lands After the Agents Already Arrived

The UK’s
Financial Conduct Authority published a wide-ranging review of artificial
intelligence in retail financial services today (Monday), the first of its kind
commissioned by a financial regulator globally, the FCA said.

The report,
led by FCA executive director Sheldon Mills and known as The Mills Review,
concludes that AI is likely to reshape how firms operate, how consumers decide
and how markets function by 2030. It does not recommend rushing to write new
rules.

Regulator Leans on
Existing Rulebook

Mills drew
on 140 written submissions and a survey of more than 5,000 UK consumers. The
review sets out four shifts it expects to reshape the sector, spanning firm
operations, consumer journeys, competition and financial crime.

The timing
is notable. A recent FM Intelligence analysis found at least ten brokers
and platform vendors wired AI agents into live client accounts in the first
half of 2026, with no regulator having written rules aimed at the practice.

Rather than
propose AI-specific regulation, the report says the FCA’s outcomes-based
approach, built around the Consumer Duty and the Senior Managers Regime,
provides the basis for the work ahead.

[#highlighted-links#]

“As
autonomy grows, the nature of regulatory risk changes,” Mills wrote.

Agent Boom Arrives Ahead
of the Rulebook

The market
did not wait for the review. In March, eToro began letting investors
delegate trades to their own AI agents inside funded sub-accounts capped at set
budgets, a feature it calls Agent Portfolios.

Robinhood
followed with ring-fenced agent accounts for its funded customers, and moomoo
added a tool that converts plain-English instructions into orders across five
markets.

ThinkMarkets
launched an agent named ChelseaAI that can
place trades but not touch client funds.

Platform
vendors joined too. Spotware opened the cTrader platform to agents through
Model Context Protocol servers, and the crypto exchange Bybit rolled out walled AI trading
accounts that keep
bot activity separate from a client’s main balance.

Most run on
the same open plumbing, an Anthropic standard released in late 2024.

The review
frames this delegation as a spectrum, with people moving from operators of AI
tools to observers who set limits and monitor outcomes.

It notes
that few parts of financial services will become fully autonomous, and that
accountability grows harder to trace as systems act on their own.

Unregulated AI Advice
Blurs the Perimeter

One finding
lands directly on regulated brokers. The report says about 26% of consumers
trust general-purpose tools such as ChatGPT, Claude and Gemini for financial
advice, despite limited awareness that formal routes to recourse will not
apply.

Roughly
nine million UK adults already ask AI apps financial questions, and only around
two in five correctly identify the protection they have when they do. The
review warns this blurs the regulatory perimeter, since tools outside it can
shape decisions without oversight.

For firms
inside the perimeter, that raises a fairness question the FCA flagged earlier
in its first technology horizon scan.

Regulated
brokers carry obligations on promotions, advice and consumer outcomes, while AI
platforms exert similar influence through a different route.

Control of the Interface
Becomes the Prize

The
competition section may unsettle intermediaries most. The report says whoever
controls the AI-mediated customer interface could gain significant market
power, and that delegation may relocate intermediation rather than remove it,
shifting it from brokers and comparison sites toward agents and platforms.

It also
warns that AI comparison could appear to search the whole market while
returning a narrower set shaped by commercial deals. Industry figures have made
similar points, with some arguing the agent could become the primary
distribution layer
between traders and the market.

Fraud and System Risk Move
Up the Agenda

FCA Chairman
Ashley Alder said the board would consider the recommendations, adding the
regulator must “keep pace with a rapidly changing environment.”

The fourth
shift covers financial crime. The review says AI will make fraud faster,
cheaper and more convincing through deepfakes and synthetic identities, and
that consumers who delegate to agents may stop challenging decisions made for
them.

It also
flags system-level risk, warning that shared reliance on a few models and
providers could produce herding and common points of failure.

In response
the FCA proposes an AI-enabled supervisory model and seven priority
recommendations, building on earlier work such as its AI partnership with Singapore’s
monetary authority.

This article was written by Damian Chmiel at www.financemagnates.com.