Regulatory requests are always unannounced and can be very disruptive for your daily business. These requests for information and documentation from regulatory bodies or external auditors come in different forms but should never be under-estimated.
In an earlier blog, we provided five tips on how you can prepare your company to limit the impact of regulatory requests. In this blog, we will focus specifically on how tools from the field of Artificial Intelligence (AI) and Data Science accelerate truth-finding missions in regulatory requests and internal investigations.
When a regulator seizes your data, they use technology to find out what happened. They want to know who was involved; who else knew what; what the damage was and who suffered. They will try to find out if and how management was involved, and if anyone did anything to cover up the misconducts.
To plan your strategy, it is crucial that you also find out what really happened. And the clock is ticking. You need to indicate what is responsive, not-responsive or privileged within a short time range (often around 10 days).
For all parties involved, it is hard to find information when you do not know exactly what you are looking for. Especially when people want to cover up or hide something.
AI and Data Science techniques like machine learning, clustering and information extraction have drastically increased the speed and improved the quality of the eDiscovery process. In handling regulatory requests, these techniques can help you to find things even when you do not know what to look for. After that, you can organize your data along typical investigation questions such as Who, Where, When, why, what, How, How Much and by Which Means.
Information extraction and text-mining techniques semantically analyze the textual data. This means that semantic entities such as persons, companies, organizations, locations, amounts, time notions, and other basic entities are automatically identified. By combining the identified entities we can detect attributes and properties, recognize relations, facts and even events. The same techniques can also detect sentiments, emotions and high-level concepts.
By extracting the right information and using the right visualizations, it is often possible to align the data based on the business application, so that the most common answers to the search questions are much easier to find and sometimes even directly clear from the data.
Even gaps in the collected email strings are identified. These gaps among suspects then can be compared and restored from email back-ups. The same is true for communication "hidden" in audio or video files, or a foreign language.
Fraudsters are not dumb: they are fully aware that all communication leaves traces. Therefore, they try to hide their communication. Encryption is not always an option, as it is very easy to detect and works as a red flag on investigators and compliance officers. They use other methods to cover up their actions by using hidden images in attachments, embedded objects or zipped information. In many cases, they also use code words; communicate by using audio files or foreign languages. As ZyLAB has been working very closely with regulatory, law enforcement and intelligence agencies for over three decades, we have seen many of these tricks. Over the years, we have developed several tools to detect and trace such forms of hidden communication and code words.