Scotiabank is increasingly turning to advanced artificial intelligence and machine learning to detect illicit financial flows and help government agencies – such as the Financial Transactions and Reports Analysis Centre of Canada (FINTRAC) – and law enforcement combat some of the world’s most serious crimes, including child sexual exploitation.
Through a series of public‑private partnerships, the Bank is applying cutting‑edge analytics to vast volumes of financial data to identify suspicious patterns that traditional monitoring systems often miss.
One initiative called Project Shadow focuses on combatting child sexual exploitation.
According to Duncan Smith-Halverson, Scotiabank’s director of AI modeling for anti-money laundering and anti-terrorist financing models and analytics, AI is not a recent add‑on but the foundation of these efforts.
“It’s absolutely critical,” he said. “I don’t think we’d be able to find as much if we weren’t doing it.”
Finding a needle in a haystack
For Scotiabank to effectively identify suspicious behaviours, the Bank had to create a system that could not just manage massive amounts of detailed information – it needed to be able to find a needle in a haystack.
In Canada alone, Scotiabank has millions of clients. Each of those clients could have anywhere from a dozen transactions a month to thousands.
The resulting datasets are vast and cannot be processed on a single machine, and require high-performance computing to manage, transform and analyze the data in a way that allows models to identify meaningful statistical patterns that could indicate potential criminal activity.
Scotiabank has procured a state-of-the-art graphics-processing-unit (GPU) cluster dedicated to anti-money laundering and anti-terrorist financing deep learning foundation models, Smith-Halverson adds. A GPU cluster works as one powerful system to perform large-scale computation tasks much faster than a single machine could.
A significant portion of the AML team’s work, Smith-Halverson said, involves ensuring that systems can “think about it all at the same time” — a prerequisite for detecting complex, multi‑party behaviour that might otherwise remain invisible.
AI leaves ‘less risk on the table’
Scotiabank and other financial institutions are required by regulation to submit financial crime intelligence through suspicious transaction reports to FINTRAC, which are also used by law enforcement. These public-private partnerships allow the Bank to work more closely and effectively with these partners.
For Project Shadow, AI enables the Bank to move beyond prescriptive rule‑based systems. Traditional transaction monitoring focuses heavily on a small set of indicators, Smith-Halverson explained, while machine learning allows analysts to identify subtler combinations of behaviour that could signal risk.
“These models help us leave less risk on the table,” he said, noting that the modeling team identifies statistical patterns while investigative teams determine whether activity should be reported to authorities.
Scotiabank is the industry lead of Project Shadow, with support from the Canadian Centre for Child Protection. The Bank works with FINTRAC, which shares the info with law enforcement agencies, including the RCMP, Ontario Provincial Police, local police forces, as well as international partners, including in Australia, the United Kingdom and New Zealand.
Smith-Halverson emphasized that Project Shadow is producing consistent, valuable intelligence. While the outcomes ultimately rest with FINTRAC and law enforcement, he said the project has significantly improved the Bank’s ability to identify patterns that might have otherwise gone unnoticed using traditional technology and approaches.
The work carries both technical and human significance, combining sophisticated analytics with investigative expertise to address a crime that is often hidden, under‑reported and deeply harmful.
Combining human expertise with powerful data analytics
Beyond Project Shadow, Scotiabank’s goal is to develop an AI model for each of FINTRAC’s public-private partnerships that it has been involved with. That includes Project Guardian, which focuses on fentanyl and synthetic opioids, and Project Protect, which targets sex trafficking in Canada. In these efforts, the Bank is experimenting with behavioural profiling techniques that help identify bad actors exhibiting similar patterns of concern.
Unlike some international trafficking models, Project Protect addresses activity that is largely domestic, requiring a deep understanding of how exploitation networks operate within Canadian financial and social systems.
According to Smith-Halverson, the project was one of the earliest public-private partnerships the Bank’s AML AI team worked on and has benefited from close collaboration with Scotiabank investigators, who bring extensive real‑world expertise to the modelling process.
That human expertise has been critical to the project’s effectiveness.
One of the investigators working closely with the modelling team previously led a Canadian police service’s human trafficking unit, providing practical insights into how traffickers operate and how financial behaviour can signal exploitation.
While most AI models remain iterative and experimental, Smith-Halverson noted that Project Protect has matured to the point where it is consistently providing valuable intelligence to law enforcement through FINTRAC, demonstrating how machine learning and investigative judgment can reinforce one another.
Despite the sophistication of the technology, Smith-Halverson stressed that human judgement remains essential. Scotiabank investigators provide rapid expert feedback that helps refine models, while AI systems increasingly play a role in translating complex numerical outputs into clear, usable intelligence.
“Trying to explain a deep neural network to an investigator by showing them the numbers doesn’t work,” he said. “Translating that into something meaningful is a huge focus for us.”
For Smith-Halverson and his team, the work carries a strong sense of purpose.
“There are a lot of jobs you can have,” he said. “Some let you feel like you’re having a meaningful impact on society. I’m very lucky to have one of those.”
