Artificial Intelligence & Machine Learning: Catalysts for Positive Change or Culprits for Malice?

Law enforcement and public safety (LEAPS) agencies, charged with upholding the law and protecting public safety, frequently rely on insights and intelligence derived from deep analysis of data. Uncovering digital intelligence can be an arduous undertaking because of the breadth, depth and sensitivity of data that is collected and reviewed by law enforcement organizations.

To make data analysis more efficient and effective, agencies are turning to artificial intelligence and machine learning (AI/ML) solutions to extract value from massive data sets in near real time. Doing so makes it possible to deliver actionable intelligence to first responders when they most need it.

AI technology accelerates analysis of investigation data, identifying and noting patterns buried in copious volumes of digital information, while ML tools draw on this analysis to arrive at valuable conclusions and predict what may happen next. They are robust tools that can help law enforcement agencies fulfill their missions. Yet, as these technologies proliferate and become more advanced, agencies must be vigilant, aware that adversaries will try to exploit AI/ML tools for their own gain.

As for use cases that apply to law enforcement agencies, AI/ML solutions would need to increase first responders’ speed and efficiency in discovering relationships that would not be possible given their time constraints and limited human resources. As such, AI/ML could be used in support of threat mitigation, predictive policing and biometric analysis (e.g., DNA, iris, facial recognition and fingerprint searches) to better serve and protect constituents. Findings from a 2022 study from the Violence Prevention Research Program at the University of California even suggested that ML tools may help identify handgun purchasers who are at high risk of suicide.

AI-powered solutions could enable law enforcement officers to search multiple disparate databases and quickly isolate criminal identifiers. They can aide in providing what is known as entity resolution. Entity resolution employs a set of algorithms to associated records from differing data sets to be attributed to a single entity—a person, place or thing.

For instance, a person with the name Michael Smyth may be recorded with different spellings of the surname—such as Smith or Smythe—or with one set of data having a date of birth, another having an address, and another having a passport number. Similarly, organizations might have multiple names. FBI Headquarters is known as the J. Edgar Hoover Building, the Hoover Building, the FBI, 935 Pennsylvania Avenue, etc. But it is one place and to associate someone or something to that place requires those disparate records be resolved. Here, AI-related entity resolution can provide the probability that records are associated.

However, as with many new technologies, use of AI/ML by law enforcement agencies has potential drawbacks. For one, advances in AI/ML technologies are of interest to criminals. Deep fakes, a form of media fabricated by AI to resemble another entity, are creating a new class of problems for agencies and private citizens alike. Digitally fabricated media has the potential to defeat facial recognition technology and breach security systems, raising concerns about the relative risks and rewards of AI-/ML- powered technologies. 

As AI/ML becomes a mainstream tool used by law enforcement agencies, questions are emerging:

• How will these technologies be rolled out to first responders and scaled to best benefit all law enforcement work? 
• Will undercover agents be aware of and on the lookout for deepfakes and other deceptive media? 
• How can biases be addressed to ensure these solutions and systems are fair in their assessments?
• How can a balance be stuck between public safety and autonomy?
• How will advancements in AI/ML affect employment and the talent shortage? 
• How can the highest quality data be selected to “feed” AI tools to ensure accurate results?
• How can the U.S. compete in the global market for the economic and national security interests that AI will contribute to?

Despite the challenges ahead, AI/ML technologies have bright futures in law enforcement and public safety organizations. Agencies should remain diligent in their AI/ML journeys and minimize the impact of human error. Successfully deploying AI/ML involves several considerations, including acquiring the right technology, training staff and avoiding cultural impediments through active change management. AI is not magic, but math that requires human management. AI/ML produces leads, but not evidence. The process for releasing new advancements must not be rushed. Agencies will seek to balance the benefits to law enforcement organizations and outweigh possible negative effects, whether unintentional or caused by adversaries. 

The potential for artificial intelligence and machine learning is unlimited. Properly used, AI/ML will help LEAPS agencies to do more than simply survive today’s dynamic threat landscape; they’ll thrive.

Mark Tanner is a board member and vice president for the Law Enforcement & Public Safety (LEAPS) Technology Forum at the AFECA Bethesda Chapter; and is president, Government Solutions and System Solutions for HunaTek.