The FCA is launching AI live testing – could this be what the financial services sector has been waiting for?

In a recent speech, the FCA asked the all-important question about AI in financial services: “So what’s holding firms back?”.

Clarity

The launch of AI live testing is about getting past some of the current challenges and helping firms make the leap from proof-of-concept to market in their AI journeys. It is about filling a current void by enabling firms to collaborate with the FCA to test and evolve AI solutions before they are released generally to the market.  

Trial period

What is proposed is a year-long AI live testing trial which, if successful, could form an established FCA service. It will form a new and additional part of the Regulatory Sandbox initiative (which includes Innovation Pathways, the Regulatory Sandbox, the Digital Sandbox and a series of TechSprints), which is wider than AI, and exists to enable start-ups, technology providers and authorised firms to trial new and innovative products and services in a safe environment. 

Explore

The live testing trial will be exploratory in nature. It will involve firms working closely and collaboratively with the FCA's regulatory and technology specialists to explore and evaluate AI solutions before they are deployed to the markets. This has the potential not only to release bottlenecks in AI deployment throughout the sector, but to influence and develop shared learning, and to inform the evolution of the regulatory landscape going forward.

Benefits

The main benefits of testing in collaboration with the FCA include the ability to explore and develop, and generate support and build understanding, in a way that will give firms comfort and confidence to invest in and eventually deploy responsible AI into the markets, with the comfort that it has been tested in conjunction with the regulator and in a designated safe space.  

Engage

The exact shape of the AI live testing has been left open for now with an Engagement Paper currently open until 10 June for stakeholders (with input encouraged in particular from chief information officers, chief AI officers, chief data officers and AI solution providers) to provide their views on three key issues:

• the current blockers to live market deployment of AI models in financial services;
• whether the live testing proposal could address the deployment challenges faced and whether it should have particular focus; and
• whether there is anything else relevant to the current situation that stakeholders want to inform the FCA about.

No new regulation 

The FCA considers that its outcomes-focused regulatory approach empowers firms to be innovative in their provision of financial services and it has been clear that it does not wish to regulate specifically for AI. The FCA considers that its current outcomes-focused regulation is clear about the regulatory expectations of firms notwithstanding the novelties of AI. However, there is a clear view from the financial services industry that further clarity would be helpful in overcoming the current caution and making it clearer to them how the current regulatory framework applies to AI.

The most recent regulatory survey (which we summarised here) demonstrates an enthusiasm for AI but evidences that significant usage can be found only in internal applications. Consequently, the financial services industry is not yet delivering direct consumer benefits through its use of AI. 

Agility and innovation

AI is evolving quickly. No-one has all the answers. Regulators need to be agile in this space. Firms need to see the regulatory requirements as foundations for the way ahead and not a barrier to a way forward. But, it seems that there is clearly some alignment of interests to reach in this space before a measurable number of firms start making the strides that the FCA would clearly like to see.

Opportunity

The FCA's AI live testing initiative could be a hugely positive step forward and provide opportunities for firms to work directly with the FCA to help to address the current uncertainties. Working with the FCA in the live testing environment would give firms the opportunity to test the performance of their AI models with support from regulatory and technology experts, and gain the all-important confidence to move from proof-of-concept to live market deployment.

The benefits from work undertaken in the live testing space could be industry-wide, with many stakeholders potentially benefiting from an evolved understanding of AI risks and mitigants, and the delivery of safe and responsible deployment, in ways that could help to drive to growth and ensure positive outcomes for firms, for the markets and for consumers.

What to expect 

The FCA has recognised that effecting the deployment of AI in financial services is no small challenge for firms and that there are a number of difficult questions to grapple with. To address some of the specific challenges that firms are facing, the FCA are proposing to explore some of the biggest questions, including how to:

• build confidence that AI-generated outcomes will meet regulatory expectations;
• assess and test the robustness of an AI model; 
• explain the output of an AI model;
• measure bias and undertake de-biasing;
• assess the impact on consumers, including vulnerable consumers; and
• address unintended consequences.

Participation

The AI live testing environment will be focused on firms that have developed an AI solution beyond the proof-of-concept stage and “are ready and willing to deploy AI models into live markets”. For firms in this position, the live testing environment could provide a unique and attractive opportunity to test and develop with the benefit of the FCA's computing power, data sets, testing capabilities and expertise in AI and regulation.

Firms wanting to participate will need to have undertaken pre-deployment testing and be ready to step forward into a controlled live market environment and to interact with consumers. Selective criteria apply and applicant firms must be able to demonstrate the following:

• that they produce or distribute financial services or products or AI models material to financial services or the financial markets;
• the active use of AI;
• the ability for the AI or technology used to drive one or more of the FCA's regulatory objectives and the potential to contribute to the UK's economic growth;
• a developed proof-of-concept and a readiness to deploy to market;
• evidence of proof-of-concept testing and market readiness; and
• a post-deployment trajectory capable of being shared with the FCA.

Still in the discovery phase?

The FCA has the Supercharged Sandbox and Digital Sandbox service offerings for those firms that are in the early stages of their AI journeys. Recent statistics suggest that “over 90% of firms that engaged with our Innovation direct support and advice services became authorised, and whilst the success rate for Fintech firms can be mixed, 80% of Regulatory Sandbox firms are still in operation. FCA Sandbox firms are 50% more likely to raise funding than their peers and on average raise 15% more in investment.”

Collaboration

All firms working with the FCA should expect to collaborate, share information and data, monitor progress and share results, and, subject to restrictions around competitively sensitive information, be prepared for wider insights to be shared with the markets.

When can we start?

Live testing is planned to commence in the summer, with two cohorts each with a six month duration, and with a report to come in spring next year detailing information on the applicant firms, the AI models tested, the application of AI to financial services, potential market impact, and evidence of positive outcomes both for consumers and for the markets.

Regulatory focus

If you would like to discuss how current or future regulations impact what you do with AI, please contact me, Tom Whittaker, or Martin Cook. You can meet our financial services experts here and our technology experts here.

You can read more thought-leadership like this by subscribing to our monthly financial services regulation update by clicking here and clicking here for our AI blog and here for our AI newsletter. 

The adoption of safe and responsible AI by the financial services industry plays a key role in supporting growth, which is why we developed the AI Lab to bring together our innovation services..... .......we believe our existing frameworks like the SMCR and Consumer Duty give us enough regulatory bite that we don’t need to write new rules for AI. And while we know from our joint FCA-Bank of England AI survey that 75% firms have already adopted some form of AI, most uses cases are for internal applications, rather than things that could directly benefit consumers and markets. So what’s holding firms back? Over the past few months, many CIOs, Heads of Innovation, and Chief AI Officers have reached out to ask? Consumer-facing applications? Can we do this? My answer is yes – the Innovation front door is always open.

https://www.fca.org.uk/news/speeches/ai-growth-how-fca-can-help