At the intersection of AI & Ethics: What Fintechs need to consider
As Artificial Intelligence (AI) begins to reshape the economy, fintechs face critical ethical considerations that extend far beyond compliance with AI regulation. The integration of AI into financial products demands a proactive approach to ethical implementation that protects consumers while fostering innovation.
For example, when AI determines credit scores or assesses loan applications, inherent biases can emerge. Whatever pre-existing biases model developers have can be inherited by the model itself. Google’s disastrous introduction of Gemini is a prime example of how it can go spectacularly wrong. Historical data often reflects societal inequalities, potentially perpetuating financial exclusion. Fintechs must rigorously test their AI models for bias against protected characteristics and implement regular auditing processes.
The challenge extends to proxy variables, which may be thought of as seemingly neutral data points correlated with sensitive attributes. Postcode data, for instance, could inadvertently discriminate against certain communities and groups if relied upon too heavily in automated decision-making.
Transparency vs. Complexity
The “black box” nature of sophisticated AI models presents a significant challenge for financial institutions. When AI influences decisions about mortgages, loans, or investment advice, customers deserve clear explanations. Yet many neural networks operate with such complexity that providing straightforward explanations becomes nearly impossible. According to the MIT Technology Review, LLMs like ChatGPT, Claude, and more recently DeepSeek, “seem to behave in ways textbook math says they shouldn’t”. Hallucinations may be less commonplace in the most modern models with chain of thought reasoning and mixture of experts, but they aren’t impossible — and that’s a problem if it affects the outcome of a financial decision.
Leading fintechs are tackling this through “explainable AI” initiatives, developing simplified interfaces that help customers understand key factors influencing decisions about their finances. This transparency builds trust and empowers customers to make informed financial choices.
Data Privacy in the AI Era
Open Banking has revolutionised financial services, but the vast data requirements of AI systems raise considerable privacy concerns. Fintechs must balance the benefits of personalised services against data protection. This means implementing privacy-by-design principles and giving customers genuine control over their data usage.
The emergence of federated learning, championed by Apple for their privacy-focussed on device AI, offers a promising solution, allowing AI models to learn from customer data without centralising sensitive information. This approach could become increasingly valuable as privacy regulations evolve.
The Evolving Customer Interface
Perhaps the biggest customer-facing change will be how customers interact with fintechs. Or to put it more accurately, how customers will interact with AI customer service agents — both chat and voice — that have been trained on the organisation’s corpus of data, calls, and emails.
UK-based tech investor Draven McConville said, “Change here is likely to come fast and given the sophistication of LLMs, customers might not even be aware of their query being dealt with by an AI-powered bot.”
Trailblazers like Octopus Energy, have trialled AI customer service as far back as 2023. Greg Jackson, the company’s CEO, wrote in The Times that the technology was “doing the work of 250 people” and “received a 80 per cent satisfaction rate, higher than the 65 per cent achieved by workers.”
Widespread adoption of AI for customer service seems almost inevitable at this point.
Financial Exclusion or Inclusion
The idea of using AI to enhance credit scoring systems is nothing new. It has the potential to both exacerbate and solve financial exclusion. While algorithmic decision-making might exclude certain groups, it can also identify creditworthy individuals overlooked by traditional metrics.
Progressive fintechs are using AI to develop alternative credit scoring methods, considering factors like regular bill payments or steady gig economy income. This could even the playing field for renters, a burgeoning cohort whose monthly rental payments have historically not been considered in credit scoring.
With Juniper Research estimating 67% growth for credit scoring services by 2028, there is a huge opportunity for fintechs to address the world’s unbanked population.
The Way Forward
The success of AI in finance will ultimately be measured not just by efficiency gains, but by its ability to improve the customer’s experience. UK fintechs have the opportunity to lead this transformation by setting high ethical standards for AI implementation.
