The rise of AI has been an ongoing and slow-burning process for a number of years now. While the technology is disruptive by nature, the recent introduction of the Gen AI ChatGPT has taken the world by storm.
Officially introduced on 30 November 2022, the technology has already made huge impacts to industries far and wide – with global news publication BuzzFeed recently revealing they would use ChatGPT to aid in its content generation efforts. ChatGPT is an example of a developing and highly promising area of artificial intelligence – Generative AI.
According to the GenerativeAI website, GenAI is the part of AI that can generate all kinds of data, including audio, code, images, text, simulations, 3D objects, videos, and so forth. It takes inspiration from existing data, but also generates new and unexpected outputs.
Despite it being common knowledge, that AI would bring huge disruption around technological and operational systems regardless of the industry, the ongoing soaring rise of Gen AI is proving a hot-rod for innovation.
Rick Grashel – co-founder and CTO of RegTech firm Red Oak Compliance Solutions – is one particular individual who understands the mammoth potential of the technology. “I am old enough that I was part of the new era of the internet in the mid-1990s,” Grashel explained. “At that time, I remember feeling that we were at the precipice of huge change. I had no idea what those changes would be, but I knew the possibilities seem limitless, and the coming change would be something on the level of the industrial revolution. When I look at Gen AI, I also have the same feelings.”
In the opinion of Grashel, the cycle of technological advancement in human history has always been the same – past technology revolutions feed into future technology revolutions – and we are once again entering a new age, he claims, not even 30 years after the previous revolution.
“Generative AI has come about because of our ability over the last three decades to collect and digitise a massive amount of our knowledge as humanity. So much of our knowledge, information, and history is available online that we can now create technology (Gen AI) that synthesizes, analyses, and creates new, highly-accurate information and content based on that body of knowledge.”
As for the impact of Gen AI on RegTech, the Red Oak exec believes the industry is going to initially see an impact in the area of content creation and moderation. “Instead of paying a team of content creators to build advertising material, financial companies will be able to provide a few basic details about a product offering to a Gen AI and then use it to create entire marketing campaigns — in seconds.
“Likewise, the reverse is also now possible. A piece of advertising material can be given to a generative AI to analyse and classify the material for compliance, categorization, and accuracy. These tasks have historically required large teams of people to perform and, with generative AI, will be accomplished by a team 10% of that size — with higher accuracy and quality.”
One of the key areas that Gen AI could help bring change in RegTech appears to center around efficiency. Muinmos CEO Remonda Kirketerp-Møller stated, for example, that Gen AI could make it easier for compliance officers to perform their duties.
“A Gen AI engine will be able to automatically prepare a prospectus, as well as fill in any accompanying reports to the regulator. Some might say this is a challenge because it may put compliance officers out of work, but in our experience, automating certain processes actually allows compliance officers to focus on other, often more interesting parts of their jobs.”
Kirketerp-Møller added that this can also dramatically improve the overall compliance quality, as tasks can sometimes be under-performed or not performed at all due to a lack of resources. “Also, regulators do not want to overload compliance teams with requirements. With Gen AI helping, regulators can boost up requirements, these requirements can be better fulfilled, thus boosting the overall level of compliance.”
With companies worldwide seeing the potential in Gen AI, the race to lead the way in this technology is already frantic and is heating up ten-fold. Jamie Hunter, chief operating officer at Aveni, outlined how the meteoric rise of AI and particularly the large language models it uses has seen the development of increasingly more sophisticated tools like virtual assistants and chatbots, which continue to transform the way people interact with machines on a day-to-day basis.
“The RegTech industry is one of those that must get to grips with this quickly,” stated Hunter. “We have seen Google emerge with its ‘Bard’ platform sooner than it might have done, in a direct response to ChatGPT – the AI platform which has amassed a staggering 100 million users in two months. AI is different to the binary technology solutions which either work or don’t. AI is constantly evolving and learning with new information being surfaced in different ways and the human element of supporting and developing this is vital.”
Beyond the changes seen in areas like efficiency, improved information retrieval, semi-automated regulation drafting and summarisations, Clausematch head of data science and machine learning Vladimir Erhsov said it is key to focus on other significant changes.
He detailed, “GAN networks responsible for the first wave of ‘this face doesn’t exist’ generated content and were trained in a generator versus discriminator fashion: essentially, two machine-learning models tried to compete with each other for generating fake images and identifying something as fake. The same idea is likely to be leveraged by large language models like Chat GPT to investigate regulations for gaps and contradictions.”
Nick Wallis – managing director for EMEA at Eventus – also cited the power of AI and ML, as evidenced by the huge response to ChatGPT, has captured the imagination of the FinTech world, which he underlined as an ‘industry always striving for new tools that can empower teams to do more faster and with higher accuracy.”
The challenges Gen AI will bring
Despite benefits around innovation, transformation and efficiency, with any new barnstorming technology – the challenges are existent and numerous.
Grashel explained, “The challenges presented by Gen AI implicate the same challenges that have arisen with the coming of the internet age — which are information sharing, accuracy, and data privacy. To be effective, Generative AI requires large bodies of accurate information to be collected and combined with prior accurate knowledge and information.’
He also remarked that financial businesses will need to become more comfortable with sharing detailed information and data about their products and assets with third-parties, who can then combine that information with other historically-known data in order to create and moderate new content.
“This blurs the lines of acceptable use, data privacy, data sharing, and data security. Financial companies will need to either relax and adapt their information security policies accordingly or risk being behind other competing companies who can create, moderate, and deliver new content more quickly,” stated Grashel.
Kirketerp-Møller also cited the fact that not every Gen AI engine is the same. She explained that it is possible that we will need to make sure that the Gen AI that produces prospectuses, for example, isn’t programmed to be issuer prone, and that its reports are not meant to portray a favourable image of the user.
“In other words, we might need to required Gen AI products used in the context of regulation to be, well, regulated – and I am a long-time supporter of regulation of RegTechs. We may need to say: “You want a Gen AI to assist your MLRO? Then just like your MLRO is certified, it has to be a certified Gen AI”.
There is also another significant challenge surrounding Gen AI, and any new technology in general – they are always susceptible to be used in a negative way by criminals and threat actors. “We know criminals are always among the first to adopt new tech and put it to use,” said Kirketerp-Møller. “This means that we might very well see a surge in fake identities, deep-fake video streams, etc., and this is something that RegTech companies will need to pay attention to and counter-measure.”
This is, however, she explained, something not only relevant to RegTech. “Publicly available Gen AI applications are already disrupting the ‘Truth Market’ by making it much easier to create fake news with seemingly reliable photos, testimonials and newspaper articles.
“As this phenomenon grows, it might be necessary to require a certain level of ‘watermarking’ of Gen AI products – something that is already being discussed among experts in the field. These watermarks will allow Gen AI produced content to be used commercially, leisurely etc., but will provide an ‘audit trail’, if you’d like, for the content, to easily rule out fake identities.”
Likewise, Hunter explained that while the outputs of Gen AI are good at the moment, the true value of what AI and LLMs can do will depend on getting specific in what is being asked and the data that is being utilised.
“This is going to be hugely important in a regulatory perspective,” explained the Aveni exec. “The data-first approach is a non-reversible trend in regulation and it must be used properly and accurately. Financial services and advisory firms are sitting on millions of data points and currently are using very little of them.”
The need for the human touch will remain key for the flourishing of Gen AI. “Data is an asset and must be identified and managed as such and this will be hugely improved with AI tools,” said Hunter. “Nuance and detail that only humans can provide will be essential, pointing the technology in a more purposeful direction through a framework, feeding it with more detailed and accurately labelled data, and developing training models to execute tasks in more sophisticated ways.”
This, he claims, will enable them to better find and use data, identify customer outcomes and lower risk, as well as make better credit-based decisions. “AI will enhance the human function and allow more impactful decisions to be made across the business.”
Ershov also underlined another key and slightly worrying change that Gen AI could bring to the RegTech space – an impact on human attention.
“It’s likely to assume that human attention will become a severe bottleneck in the age of the AI era. It’s already a noticeable weakness for the system. For example, think of FAANG companies and others placing bots as the first line of defence already. But once bots are on both sides, it would be impossible for humans with a custom case to reach out to the real human at the other end to get his attention.”
The undeniable watchword when it comes to understanding the role of AI in financial compliance is limits. Wallis explained that regulators globally are asking that companies have systems with ‘explainability’, as they noted many AI/ML systems struggle in this area.
He explained, “Regulators are looking for more transparency whether an AI/ML model is generating an alert, a string of text, or identifying trading behaviours. For AI systems, there are also well-known methodology problems with overfitting and bias that require a team of highly sought-after engineers to continually re-tune. This is an advanced technology that, when used, requires continual maintenance.
“Instead, we expect RegTech is best served with methods that blend rules-based, ML, and automation and do so with clear documentation and auditability at each step.
The future for Gen AI
At a strategic level, AI can, and should be a game changer for the risk function to make it less operational and a clear value driver – claimed Hunter.
“The ability to generate human-like speech, text, image and code opens up new possibilities for making tasks significantly more efficient or automating them completely. Through technology and budgetary opportunity, the risk function must utilise valuable data that can impact the bottom line and derive greater value for customers and shareholders.”
While the discussion surrounding Gen AI will continue to evolve, there are some who believe RegTech is in the right position to be a leading voice on the matter. “I believe that RegTech, which is positioned in the crossroad of law (regulation) and technology, is actually a most suiting arena for such discussions and dilemmas, and it’ll be good to see more RegTech stakeholders, such as regulators, take point on this in the coming future,” concluded Kirketerp-Møller.
Anthony Quinn – CEO of Australian RegTech firm Arctic Intelligence – also commented that AI is and will continue to have a big impact on how firms manage regulatory compliance obligations.
“AI has the power to help in many ways, such as analysing huge data sets to identify patterns, trends and even answers that might take hours to investigate manually. The opportunities for time savings could be massive but we should expect challenges too, such as validating the accuracy of the results and any hidden or subtle bias that may exist in the construction of AI based models.”.
Grashel mentioned that for financial and RegTech firms, acknowledging the presence and impact of Gen AI is not enough. “These companies should put together initiatives to start thinking about how this technology implicates their own businesses and plan accordingly. Otherwise, they risk being passed by other forward-thinking, progressive companies — or worse — becoming irrelevant.
“Generative AI is here. Not unlike the arrival of the internet age, I believe it will change everything. The only question is how quickly and proactively RegTech and financial services companies adapt to deal with it.”
Copyright © 2023 FinTech Global
Copyright © 2018 RegTech Analyst