I’m not sure if it’s still the case but when I worked at the FCA Wednesday was always an important day. Not just because it’s the middle of the week, but because it was fish finger sandwich day at the hot deli counter in the staff canteen. It’s the small things that keep regulators happy.
This particular Wednesday, at the time of writing, we’re waiting to read the latest speech by deputy Chief Executive Sarah Pritchard, speaking at the Investment Association. It’s one of multiple speeches from senior FCA people this week, starting with chief data, information and intelligence officer Jessica Rusu on Monday about implementation of AI, alongside its report on the good and not so good examples of how to do it.
Her speech was mostly about how the regulator is implementing the recommendations of the Mills Review of AI in financial services. It’s taking forward six of the seven recommendations, but as is often the case with the regulator, it’s most interesting to look at the one it isn’t taking forward – the development of a public-interest financial capability tool.
The Mills Review wasn’t the only publication on the regulation of AI to come out over the summer. HM Treasury produced its Financial Services AI Adoption Plan around the same time, written by industry-appointed “AI Champions” – Harriet Rees (Group CIO, Starling Bank) and Dr Rohit Dhawan (Head of AI, Lloyds Banking Group).
One key difference between the approach of the two publications was that the Plan, it will not surprise you to learn, concluded that a more industry-centric approach is a better way forward. Its answer to the advice gap problem for example, is not a public-interest robot but for the industry to voluntarily agree a consumer disclosure label distinguishing regulated from unregulated AI output. Watch the lang cat’s space shortly for more research on how that might pan out, given current consumer perceptions of whether AI information is or isn’t regulated.
Given that we’re not getting a robot from the FCA it’s therefore up to the industry to make things work. But the other recommendations from the Mills Report are being taken forward, meaning the regulator will still be actively engaged. These are securing and adapting the regulatory perimeter, strengthening system-wide coordination and oversight, monitoring the transition to autonomous models and adapting regulatory frameworks. The FCA itself is scaling up its AI Lab to support AI models and system innovation in financial services. It’s planning to work with industry, other regulators and frontier labs to develop a trusted framework for agentic finance, which will attempt to provide clear approaches to agent identity, delegated authority, accountability and execution.
We can expect to hear more about all this at Nikhil Rathi’s Mansion House Speech on Thursday. Last year he used the opportunity to deliver a sharp rebuke to the industry for not having sufficient cyber security processes in place, and that’s still very much on the radar, underlined by the issues with Google Cloud yesterday.
And last, but not least, FCA Chair Ashley Alder is speaking about international regulatory cooperation in Washington on Friday, underlining the point that AI pays no attention to national borders.
I hope they do still have fish finger sandwiches in the FCA deli – fish is supposed to be good for the brain, and it looks like they’re going to need it.
For your music choice, if you’re reading this on Wednesday, happy 69th birthday to Julian Cope, and a huge slice of 80’s nostalgia as he careers around in a Jeep with a noisy brass section for Reward.
