Regulation
FCA publishes good and poor practice on AI, including governance layers and an inventory for AI agents
On 19 October 2026 the FCA published illustrative examples of good practice and areas for improvement in firms' use of AI. It says the examples create no new requirements, and it singles out agentic AI as likely to need stronger controls. The page has no trading-specific guidance.
On 19 October 2026 the Financial Conduct Authority published 'AI in financial services: good practice and areas for improvement', part of its good and poor practice series. The page was updated on 20 October with an editorial amendment to one chapter. The FCA states that the examples 'do not create new requirements'. It points firms back to rules that already apply, including the Principles for Businesses, the Consumer Duty, the Senior Managers and Certification Regime, systems and controls rules and the operational resilience rules, and it says firms remain responsible for judging what is proportionate for their own use of AI.
Agentic AI gets specific mention. The glossary says it 'may need stronger controls because it can operate more independently'. Under good practice, the page describes building governance layers for agentic AI before more autonomous systems are deployed, and registering agents in a central inventory so that a firm knows what each agent can access and who owns it. This follows the FCA's speech of 19 October, which set out the idea of a wider framework for agentic finance without a timetable.
The rest of the document covers ground that applies to any firm using AI. Governance should be proportionate and risk-based across the whole life of a system. Senior managers are expected to take reasonable steps to understand AI risks, but the page says the FCA does not expect them to be technical specialists. Monitoring should track real-world outcomes rather than only technical measures, and reactive or static monitoring is listed as an area for improvement. Human-in-the-loop controls should be risk-based, and the page warns against excessive reliance on them turning into a box-ticking exercise. On resilience, firms are told to avoid overreliance on a single model or vendor and to plan for model changes and exit options, and that they remain responsible for outsourced AI services.
For anyone running or building a trading agent, the relevance is indirect. The page does not contain examples or guidance specific to trading, investment or market-facing AI; its examples centre on consumer journeys, governance and operational resilience. It is also aimed at authorised firms, so it is not a rule for an individual using an agent with their own money. What it does show is the vocabulary a UK regulator is using when it asks about agents: who owns each one, what it can reach, how its outcomes are watched and what happens if a model or vendor changes. Those are useful questions to be able to answer about any automated trading set-up.
Two limits apply. First, apart from the speech mentioned above, this briefing relies on the FCA page alone; it has not read the linked rules or the supervisory and testing work the page names as sources, so it makes no claim about what those contain. Second, the document sets out illustrative examples and no deadlines or enforcement commitments, and nothing here is advice or changes what an individual may do with a trading agent today.
Read the original: Financial Conduct Authority (AI in financial services: good practice and areas for improvement) - FCA publishes good and poor practice on AI, including governance layers and an inventory for AI agents. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.