Artificial Intelligence is transforming every industry and area of business. But much like modern technology in general, it is a double-edged sword. On one hand, when used correctly, it can bring efficiency, automate processes, and reduce costs. On the other, it provides bad actors with a powerful new tool for misconduct. Financial crime is no exception.
In front-line AML Operations, an ever-increasing number of AI tools and machine-learning models operating on behavioural patterns are enhancing manual processes or even replacing them completely. For example, transaction monitoring and sanctions screening alert processing are among the most time-consuming tasks for AML Analysts due to the large number of false positives.
In the crypto space, AI algorithms can analyse millions of blockchain transactions to identify wallet addresses that may belong to the same criminal network. These tools can detect behavioural patterns, trace the movement of funds, and uncover hidden relationships between wallets at a scale that would be practically impossible through human review alone.
In KYC (Know Your Customer), the introduction of biometric checks to verify the identity of customers has had a huge impact on the onboarding process. What once took days now takes minutes: a customer takes a picture of their identity document, the system checks it against a live selfie or video, and verification is completed in real time.
For non-retail customers, extensive corporate documentation (articles of association, annual financial statements, etc.) can be instantly translated, summarised and analysed.
AI tools can also be used to streamline less operational tasks, such as supporting the drafting of Suspicious Activity Reports, updating internal policies and procedures, and carrying out regulatory impact assessments.
Why, then, are there still trillions[1] of dollars estimated to be laundered every year? Partly because practitioners aren’t the only ones making use of AI. Criminals are too, and some of them are ahead.
Deepfakes are easily the biggest threat in identity fraud. AI-generated identity documents, voice cloning tools and synthetic faces are easily accessible, increasingly convincing and being used to commit fraud across the globe. In biometric KYC checks, we see the same technology used to enhance customer onboarding being used to manipulate those same controls. Voice cloning is becoming a serious threat in payment fraud, where victims are manipulated into transferring funds to criminals believing they are sending them to a family member or friend in need, falling victim to romance frauds, or “pig-butchering”.
And whilst AI can bring significant benefits, these tools could do more harm than good if not used correctly. They require appropriate calibration, controls, and regular testing, and cannot be relied upon for high-risk or complex cases. Data protection must also be carefully considered, particularly regarding the risk of third-party data sharing when inputting personal customer details into AI systems, and the regulatory limits on automated decision-making. When it comes to Financial Crime, Human-in-the-Loop is not enough. The consequences of getting it wrong range from poor customer experience (a customer being unable to verify their identity using biometrics due to a system error, for example) to significant fines and reputational damage resulting from AML failures. Perfection is not expected from the regulators, and the same way that they distinguish between human error and negligence, they are likely to distinguish between machine error and inadequate oversight. What they do expect from regulated financial institutions is a robust control framework and clear evidence of human decision-making.
It is precisely for this reason that financial institutions should be cautious when reducing their AML workforce; instead, what we should see is an evolution of roles where manual processing decreases, and more emphasis is put on oversight, system calibration, and testing. Still, entry-level professionals are undeniably those in the greatest danger of losing their jobs. On the other hand, subject-matter experts (such as sanctions professionals, crypto investigators, and complex entity specialists) are more resilient to automation. While AI is likely to automate significant portions of compliance work, it is difficult to envisage a future in which the Compliance Officer or MLRO role disappears entirely. Accountability and personal responsibility are foundational to these positions and assigning criminal or regulatory liability to an algorithm simply isn’t going to happen.
The question isn’t whether AI will impact financial crime compliance – it already has, positively as well as negatively. The question is whether the industry will manage it responsibly to stay ahead.
Sources:
ACAMS Today (2024), The Use of AI and Machine Learning in Financial Crime Compliance, https://www.acams.org/en/opinion/the-use-of-ai-and-machine-learning-in-financial-crime-compliance
European Banking Authority (2025), Opinion on the risks of money laundering and terrorist financing affecting the EU's financial sector, eba.europa.eu
EU AI Act. https://artificialintelligenceact.eu
FATF, Horizon Scan AI and Deepfakes (2025). https://www.fatf-gafi.org/en/publications/Methodsandtrends/horizon-scan-ai-deepfake.html
[1] The estimated amount of money laundered globally in one year is 2 - 5% of global GDP, or $800 billion - $2 trillion in current US dollars. unodc.org/unodc/en/money-laundering/overview.html
