Featurespace leverages upside of generative AI to deliver hammer blow to online credit card fraudsters

Cambridge FinTech pioneer Featurespace is harnessing the positive deployment of generative AI to combat escalating $multibillion online credit card swindles after inventing a first ever transactional model of its fraud-busting technology.

TallierLTM stymies scammers by using a language they don’t understand – figures rather than words – which puts control firmly in the hands of clients’ data scientists.

The technology is being launched in conjunction with long-term Featurespace client – Wall Street-quoted global payments company TSYS.

While created for clients globally, America is set to be hit by around 40 per cent of the world’s $397.4bn of card fraud losses in the next decade.

Harnessing 10 years specialisation in the use of AI for good, Featurespace’s co-founder David Excell says that “What OpenAI's LLMs have done for language, TallierLTM™ will do for payments.

Recently feted as a major winner in a joint UK-US government Privacy Enhancing Technology (PETs) programme to discover new ways to combat online fraud, Featurespace listened to the concerns of its clients and the growing need to protect customers in the face of fraud and came up with this gamechanger.

TallierLTM is the world's first Large Transaction Model – hence the LTM and is a foundation AI technology for the payment and financial services industry.

It is a large-scale, self-supervised, pre-trained model designed to power the next generation of AI applications for the financial protection of consumers. TallierLTM has shown improvements of up to 71 per cent in fraud value detection when compared to industry standard models operating at an industry-typical 5:1 False Positive Ratio.

Featurespace says the technology will provide a significant improvement when it comes to differentiating between genuine consumers and bad actors. Seventy per cent of financial institutions in North America consider financial criminal attacks to be getting worse, with exponentially increasing losses, as scammers become more sophisticated.

Featurespace co-founder David Excell said: “What OpenAI's LLMs have done for language, TallierLTM™ will do for payments. “There is widespread concern about how deep-fakes and generative AI have been used to deceive consumers and our financial systems.

“We plan to reverse this trend by utilising the power of generative AI algorithms to create solutions that protect consumers and make the world a safer place to transact.”

TallierLTM has been pre-trained across jurisdictions and market segments using a self-supervised approach, making it highly accurate and representative of real-world consumer transactions.

By analysing billions of transactions, TallierLTM identifies hidden transactional patterns undiscoverable using current industry methods, enabling it to generate likely future consumer transaction patterns.

Insights are based on time sequencing, such as unusual spending patterns over a short period and patterns of behaviour between a consumer and a merchant. These are, for data scientists, a critical task when differentiating between genuine and criminal activity.

Financial institutions will be able to interact with TallierLTM via its embedding API, which is a data science accelerator that enables a consumer’s transaction history to be converted to a machine-readable feature vector.

It creates a unique 'behavioural bar code,’ providing a comprehensive representation of a consumer’s transactional behaviour without revealing any personally identifiable information.

Dr David Sutton, Featurespace's chief innovation officer, said: “We know that smarter technology helps financial institutions better understand their consumers.

“We have taken this to the next level by pairing cutting-edge generative AI algorithms with huge volumes of data, enabling a machine to efficiently comprehend the relationships between different customer transactions.

“By adding TallierLTM's feature vectors to the inputs of an industry-standard fraud model, we've seen improvements of up to 71 per cent in fraud value detection.

“This will accelerate data science teams' ability to level up their model performance and realise the value of machine learning investments more quickly.”

Dondi Black, executive vice-president and chief product officer for TSYS said of the partnership: “TSYS and Featurespace have a long track record of co-innovating to deliver industry-leading technologies and solutions that help financial institutions provide secure, seamless experiences at scale.

“TallierLTM is the next transformative step, allowing us to unlock new insights and power new experiences for our clients around the world, further differentiating our capabilities with this new groundbreaking solution.”

A new peer-reviewed scientific paper detailing the advances that enabled the development of TallierLTM™ has been accepted to appear at the 2023 ACM (Association for Computing Machinery) International Conference on AI in Finance in late November in New York.

• Companies interested in learning more about TallierLTM™ can register their interest at https://www.featurespace.com/tallierltm/