Three Ways Generative AI Can Help Criminal Investigations

Image: Franz P. Sauerteig from Pixabay

Author: Ellen Joyner-Roberson, Global Marketing Advisor at SAS

Around the world, investigators and law enforcement officers are faced with a rapidly changing and increasing workload, driven by an increase in complex, modern-day crimes.

Because technology is changing the type and methodology of crime itself – tax evasion, theft of government funds, wrongful payment of benefits, abuse of government systems, property crimes, and heinous crimes against people – it can also help investigators.

Researchers use machine learning (ML) models to detect activity that matches identified fraud schemes and to identify new and emerging patterns and trends. In addition to detecting established crime schemes, ML can find suspicious transactions when criminals change their schemes. Researchers also use natural language processing (NLP) and text analytics to extract information from social media, crime reports, and other documents.

GenAI is the latest technology that can help investigators assimilate information to make better decisions and work more effectively. A recent fraud technology survey by the Association of Certified Fraud Examiners (ACFE) and SAS found that 83% of fraud investigators expect to add GenAI to their toolkits within the next two years.

Below are a few examples of how traditional AI, augmented with GenAI, can help researchers.

GenAI as a digital assistant for special investigations

Crime investigations typically involve a vast amount of intelligence reports. These reports are extremely time-consuming to read, but extracting key people, addresses, phone numbers, and relationships that are relevant evidence to a case can also be particularly difficult. New information extracted from a crime report requires reviewing previously read reports, making the process repetitive and time-consuming.

GenAI and large language models (LLMs) can help find information items and connect the dots between them faster and easier. An LLM-powered “digital assistant” can be invaluable because it interprets intelligence data to answer questions and extract the most relevant information. A digital assistant can generate summary stories, highlight crucial details, identify potential gaps and conflicts within the investigation process, and suggest next steps.

What sets this approach apart is the adaptive nature of GenAI. It learns and evolves with user feedback, continually refining its models and providing deeper contextual understanding within the research domain. This dynamic interaction ensures accuracy, explainability, and transparency at all points in the process.

In addition, the collaboration between LLMs and traditional AI improves many other aspects of crime and investigation work.

GenAI for conversation analytics

Conversation analytics has redefined the way investigators approach digital exchanges. The latest advanced tools are revolutionizing investigation risk assessment by capturing and organizing transcripts of digital exchanges on mobile devices. These solutions present the data in an easy-to-use viewer that displays the exchange between two or more people.

The tool’s main feature is the ability to navigate and select highlighted key terms, which speeds up the process of identifying opportunities in massive logs of conversations. This can be useful in crimes such as online child sexual exploitation and online grooming, where a conversation between a predator and a victim reveals evidence of manipulation, an exchange of money, or a meeting place. It can also identify risk indicators in social media exchanges, such as escalating behavior within the conversation.

The addition of GenAI capabilities to the conversation analytics tool enables law enforcement to more closely examine large chat logs and find evidence of concern, ongoing criminality, and threats of risk and harm hidden in digital exchanges.

GenAI for Post-Incident Assessments

GenAI can also make the job of dispatchers easier. An agent receiving a call must obtain as much information as possible from the caller, articulate the incident in the system, and identify risks, actions, and resources accordingly. This high-pressure task is prone to human error and personal perspective. Through procedural documentation, LLMs can review dispatch data and decision-making, providing an opportunity to support subsequent decisions and ensure that the quality of service in an organization can be maintained.

A look into the future

The use of AI and GenAI marks a significant milestone in the development of investigative practices in various crime domains. This technological synergy offers investigators an opportunity and a transformative possibility to modernize their activities, leading to more effective crime prevention and detection strategies.

SAS Analytics Week – Public Safety Editionwhich takes place from August 13-15is a pivotal event that brings together experts and industry leaders to explore how technology and analytics can improve public safety. With topics ranging from preventing gender violence to combating online child exploitation, this event not only highlights the importance of these issues but also provides practical, data-driven solutions. Find out more and register here.