AI · 21 October 2026
FCA Proposes AI Failure and Close-Call Reporting for UK Firms
The UK Financial Conduct Authority wants financial services firms to report AI failures and close calls as banks deploy more autonomous AI agents, according to The Register. The proposal focuses on accountability and catching problems before they cause lasting harm.
What happened
The UK finance regulator, the Financial Conduct Authority (FCA), wants financial services firms to report AI failures and close calls, according to coverage in The Register. The push centres on establishing clear accountability as banks and other financial firms deploy increasingly autonomous AI agents.
The reporting proposal treats near misses as well as outright failures, so that problems can be identified before they cause lasting harm. The reporting excerpt available to us is brief, so the detail on scope, thresholds and timing should be checked against the regulator's own publications.
Why it matters
Autonomous agents make decisions and take actions across multi-step processes, often with less human review at each stage. When one goes wrong, the effect can spread through a customer journey before anyone notices. A requirement to report close calls shifts the focus from proving that failures never happen to learning from the moments when they almost did.
For leaders in experience, AI and digital transformation, the signal is that accountability for an agent's behaviour must be designed in from the start. Who owns the outcome when an agent acts on a customer's behalf? How quickly can an organisation see that something is drifting? These questions sit with operating-model design as much as with compliance.
The Renascence take
Most of the conversation about this will be framed as a compliance matter. That misses where the risk actually lands, which is in the customer experience.
Close calls are early signals of experience debt: the mismatched refund, the wrongly declined payment, the confusing hand-off that customers absorb before anyone logs it. The principle underneath is that people judge a service by its worst recoveries, not its average transactions. Operators should map every agent decision point to a specific customer journey, name one accountable owner for each outcome, and feed near-miss data into service design reviews rather than filing it away for regulators alone.
Sources
This briefing was written by our Newsdesk, synthesising reporting from the outlets below. Follow the links for the original coverage.
FAQ
Questions we get on this topic
More in AI
Stay ahead of CX
Get the signal, not the noise.
The stories shaping customer experience — plus the Journal and Experience Loom — in your inbox.


