Why financial regulation needs a different kind of AI

Regulation is often portrayed as an obstacle to innovation, yet its purpose is quite the opposite. Well-designed regulatory frameworks exist to protect consumers, maintain confidence in financial markets and create the conditions for new products and services to emerge responsibly. The challenge is that as financial services have evolved, the body of regulation surrounding them has grown steadily more complex. Every new framework addresses a legitimate need, but together they have created an environment that is becoming increasingly difficult to navigate, regardless of where a business operates.

The pace of change has looked different from market to market, but the direction of travel has been remarkably similar. Europe has introduced successive frameworks including MiFID II, GDPR and the Digital Operational Resilience Act (DORA). In the US, the Dodd-Frank Act reshaped financial oversight following the 2008 financial crisis. Brazil has followed a similarly ambitious path as Open Finance, Pix, crypto and other digital payment initiatives have matured, each bringing new regulations, guidance and reporting requirements.

As regulation expands, so does the cost of understanding it - and that shift is beginning to influence innovation in ways the industry has yet to fully address.

The hidden cost of regulatory complexity 

For many businesses, the greatest challenge is interpreting the data correctly. Establishing whether a product meets regulatory requirements often involves working through thousands of pages of legislation, guidance and technical documentation before development can even begin, a process that can demand significant legal and compliance expertise.

For smaller teams in particular, the resources needed to navigate that complexity can become a barrier in themselves. If the route to compliance depends on access to specialist expertise, fewer businesses are able to bring new products to market. That ultimately limits both innovation and competition, as promising ideas can be held back long before they ever reach customers.

The result is an uneven playing field. Organisations with established compliance functions can move forward with greater confidence, while smaller teams are often left interpreting the same rules with far less certainty. Delays, unnecessary caution and costly mistakes become more likely, widening the gap between those with ready access to regulatory expertise and those without.

The limits of general-purpose AI

When faced with a complex problem, most people do what has quickly become second nature: they ask AI. According to OpenAI, more than 200 million users ask ChatGPT financial questions every month, while 65% of people use AI chatbots for legal advice. It is reasonable to assume many of those questions relate to professional and regulatory challenges, not just personal matters.

For many tasks, general-purpose AI is remarkably effective. Financial regulation, however, demands something different. Widely available models are trained on broad pools of publicly available information, making them well suited to summarising, explaining and generating content. They are not designed to interpret complex regulatory frameworks or provide guidance tailored to the requirements of a specific regulated product or service.

Accuracy is only part of the equation. In regulated industries, users also need to know where an answer comes from and whether it can be trusted. McKinsey found that almost a third of organisations have experienced negative consequences associated with AI inaccuracies, while more than half are actively working to reduce those risks. When compliance is at stake, a plausible answer isn't sufficient; you need the right answer. General-purpose AI can be a useful starting point, but without specialist knowledge and a reliable way to verify its responses, it’s impossible to have the level of confidence that regulatory interpretation demands.

A different approach to regulatory AI

Applying AI to financial regulation requires a different foundation. Regulatory AI should draw on official documentation, incorporate regulatory updates as they are published and reference the guidance behind every response.

That becomes particularly important in financial services, where a single regulatory framework can encompass thousands of documents. For engineering and product teams, navigating those requirements can take weeks before development even begins. Focusing on specific domains, such as Open Finance or instant payments, allows a model to operate with greater depth while linking every answer back to its original source.

Brazil offers an early example of what this approach can look like with Regulus, an AI chatbot grounded exclusively in official Brazilian sources — the Central Bank's Sisorf manual, the Pix rulebooks and the documentation published by Open Finance Brasil. Instead of returning a standalone answer, it directs users to the regulation behind its response, making it easier to understand how that conclusion has been reached rather than hallucinating regulatory details, confusing regulations or mixing up jurisdictions.

We built Regulus to solve the same challenge our own engineering team faced when building products within Brazil's Open Finance ecosystem. Rather than attempting to cover every area of financial regulation, it deliberately prioritises depth over breadth, helping developers and product teams interpret regulatory requirements before designing and implementing new features. If Regulus cannot find the relevant information, it says so explicitly rather than generating an answer.

It is no coincidence that this approach has emerged in Brazil as the country has one of the world's most advanced Open Finance ecosystems - yet many businesses still face practical challenges navigating an increasingly complex regulatory landscape. That makes it an ideal environment for exploring how specialist regulatory AI models can improve access to regulatory knowledge without compromising accuracy or accountability.

Lowering barriers to innovation

Financial regulation will continue to evolve alongside innovation, but access to regulatory knowledge must evolve with it.

General-purpose AI has transformed how people find and use information. In highly regulated industries, however, broad knowledge alone from a publicly available model lacking domain expertise is not enough. Trust depends on authoritative source material, continuous updates and the ability to trace every answer back to the regulation it is based on. Drawing on a single jurisdiction's official regulatory information, with live document indexing and mandatory source citation, could provide a solution that supports healthy competition and protects companies from being led astray.

Removing barriers to regulatory expertise has the potential to strengthen the entire financial ecosystem. It gives smaller businesses the confidence to build and launch new products, allows established organisations to work more efficiently and helps ensure regulators spend less time correcting avoidable mistakes. As technology continues to reshape financial services, improving access to trusted regulatory knowledge may become just as important as the innovations themselves.

Editor's note

Every piece published on The Bright Minded goes through careful verification, but mistakes can happen. If you spot an error, have additional information, or want to flag anything, write to rosalia@thebrightminded.com.

This is NOT a sponsored article. The Bright Ones lets fintech people talk about their projects and the reasons why they improve people's lives and the industry.