AI for protecting against future threats

When: Already a problem

Bad actors often innovate faster than good actors. As AI capabilities advance, criminals are armed with a new set of tools and opportunities to attack with higher frequency. Generative AI is enabling more sophisticated and more creative threats than ever. 

Future threats posed by AI include:

• Theft—direct attacks on banking infrastructure to steal money/data

• Impersonation—use of AI agents to impersonate people and attack weak security

• Social engineering—AI deepfakes and other social engineering tactics to scam customers

• Fraud—AI generated synthetic data to circumvent AML, KYC and authentication controls

So how do Financial Services companies defend against these attacks? 

The only way to respond to this evolving threat is to upgrade defences by deploying the best AI experts armed with the latest tools. Fintech startups will have a huge role to play in this space due to their ability to attract the best talent and build using a greenfield tech stack, whilst incumbents have the huge data lakes required to build the LLMs required. 

Fintechs will help us detect deep fakes and protect consumers from scams—such as platforms like WILTY that can help you scan photos or videos and identify them as deep fakes. For older generations, who are typically at higher risk of being scammed, fintechs like Carefull help protect your money, credit, and identity from threats. 

There’s significant demand for more accurate fraud detection. Banks spend billions on fraud detection but suffer from models which are over-sensitive with false positives costing  merchants around $400bn in value last year—10x the value of genuine fraud! This is largely due to a lack of contextual data resulting in models being trained on data that is biased towards good transactions and overcompensating. 

Mastercard are demonstrating how incumbents can lead the way here, developing a new AI model built on a neural network that provides better context around whether the person is likely to spend money at the merchant based on their history, profile, etc. 

Insurance companies like Aviva are also turning to AI to reduce fraud. LLMs using AI image/video recognition can be used to identify legitimate claims and reduce the time it takes to settle these claims.

AI finance for AI natives

When: 5-10 years

Gen Z has been labelled the first ‘internet-native’ generation, having grown up surrounded by and immersed in the internet. The next generation (Alpha) are likely to be dubbed the first ‘AI-native’ generation. Where Millenials and Gen Z may use ChatGPT to replace search and Midjourney to power memes, Gen Alpha will have an AI agent for all everyday tasks, have AI friends, and be taught by AI teachers.

They will likely see fundamental changes in the way they work and the way they consume products. And as this generation comes of age, they’ll start engaging with financial products in radically different ways. 

So what financial services does a consumer need when they’ve grown up in a world of AI?

Much of this will come down to interface and the way financial services look. A generation more comfortable with conversational interfaces might see a shift away from typical banking app interfaces and towards conversation, sending and receiving payments via message, for example (AKA Venmo for everything). 

There’s of course a growing literacy and comfort with new digital asset classes, and AI could be the vehicle to drive the shift towards ownership of virtual homes, virtual land, virtual apparel (which crypto has not been fully able to deliver on). 

Infinite optionality and ultra-customisation seems a likely demand. While we currently choose between competing financial products across different providers, AI natives will likely look for an infinite number of products and resources at the click of a button. Financial services can’t continue to exist with a small selection of products—they’ll need to expand, diversify, and adapt to the ever changing needs of their users. There are some fundamental questions that need answering here. How do you even start going about building a bank that serves this?  

Personalisation may have also advanced so far to the point that the algorithms you use are so predictive, you don’t even need to think or execute decisions—they’re just made automatically.