By this point , it is becoming clear to most banking institutions and compliance officials that technology can be a potent weapon to be used against financial crimes , such as money laundering and fraud . A recent artificial intelligence ( AI ) adoption survey revealed that nearly half of the compliance professionals surveyed had allocated some portion of their budget to AI projects , with close to 20 % already utilizing AI and machine learning models to thwart malicious actors . 1
According to a 2022 study by LexisNexis , the true cost of financial crime compliance has been on the rise since the pandemic , with global costs estimated at $ 213.9 billion . The U . S . alone experienced a 33.3 % compliance cost increase between 2019 and 2020 . A key component of the increase has been due to increased investment in labor — with labor ’ s percent of the share of compliance costs moving from 54 % to 59 % between 2019 and 2020 . 2 Given the rise in compliance costs , many financial institutions ( FIs ) are looking into innovative ways in which technology can provide a significant reduction in costs by creating a more productive and efficient environment for investigators .
At present , a new layer of AI is swiftly emerging that promises to significantly enhance the detection of financial crimes . These “ generative AI ” and “ large language models ” have been making significant waves across various sectors . Let us delve into what these terms mean , how they are employed in distinct industries and , particularly , how they could aid in the investigation of financial crimes .
What is generative AI ( GenAI )
To begin , GenAI generates original content such as images , text and music , emulating human creativity . It recognizes patterns and rules from extensive data sets and produces new content accordingly in record time . For example , an AI model trained in landscape images can show you realistic images of what your backyard could look like if it had a pool .
Despite the ongoing buzz around GenAI , it is far from a novel concept . Indeed , GenAI has its roots in the 1950s , with early iterations generating sequences of data like text or speech based on probabilistic patterns . Nevertheless , recent significant advancements in deep learning , fueled by the progress made by OpenAI , like the ChatGPT models , have propelled GenAI to the forefront , holding the promise of increased utility .
ACAMS Today September – November 2023 91