Artificial Intelligence at the FBI – 6 Current Initiatives and Projects

Millicent Abadicio

Millicent is a writer and researcher for Emerj, with a career background in traditional journalism and academic research.


The Federal Bureau of Investigation has figured in many works of fiction as well as documentaries because what they do fascinates people. The representations have not always been favorable, but most people agree they have spearheaded many advances in science and technology in law enforcement.

It seems that either that has not been the case when it comes to artificial intelligence or the FBI is keeping their AI projects mostly under wraps. If they have fallen behind on that front, it’s perhaps understandable. As a federal agency overseeing domestic crime and investigations as well as foreign threats within its borders, it has to answer to a whole slew of oversight committees for everything it does, such as using facial recognition software. The FBI cannot simply use technology as a commercial company such as Facebook would.

Nevertheless, AI at the FBI has been gathering steam in some areas of law enforcement. In fact, senators from both parties collaborated on a bill, the AI in Government Act of 2018, that would look more closely into the potential of AI technology for government use. If it passes, it could pave the way for developing AI at the FBI.

At around the same time, the FBI secured funding that would enable them to secure AI talent for their field offices, primarily for data management, but possibly for other areas as well. Current AI applications that could benefit the FBI include:

  • Facial recognition
  • Fingerprint identification
  • DNA matching
  • Cybersecurity
  • Insider threat identification
  • Business process management

We discuss each in detail below, starting with facial recognition.

Facial Recognition

Facial recognition software is taking off in the commercial sector, and some of the biggest companies in banking and finance are reaping the rewards.

Many people have heard about the Integrated Automated Fingerprint Identification System (IAFIS), perhaps from watching true crime television shows. The FBI, through its Criminal Justice Information Services (CJIS), developed it over many years, and it has been a great boon in criminal investigation.

However, with big data comes great responsibility, so the FBI turned to AI to develop the Next Generation Identification (NGI) technology. One of its capabilities is comparing images to identify those associated with criminal activities using the Interstate Photo System database. Below is a short 1.5-minute video showing how the NIG-IPS system works:

The FBI use of facial recognition software has had its successes, but some people are not convinced they should use it due to civil liberty issues.

The FBI is also looking into commercially available software for video surveillance, particularly Rekognition by Amazon. While this is still exploratory, this move has come in for its share of criticism from numerous advocacy groups.

Fingerprint Identification

The FBI has been using IAFIS primarily to match fingerprints from a known ten print database, and the NGI has expanded that capability. The new system, known as the Advanced Fingerprint Information Technology (AFIT), also looks at latent palm prints.

This 10-minute video shows how law enforcement in Texas was able to use AFIT to solve an actual case in Texas from time mark 2:51 to 3:24.

However, the FBI has frequently encountered cases of obliterated or otherwise changed fingerprints that prevented identification. To circumvent that, the agency has put out requests for information about AI technology that might be able to make a match despite these intentional or unintentional alterations.

Other capabilities of NGI include:

  • Repository for Individuals of Special Concern (RISC) – this is an early program for identifying wanted and other dangerous individuals using a fingerprint database; designed to provide law enforcement personnel with on-scene and instant access via a mobile device
  • Friction Ridge Investigative File – contains all relevant images and events associated with an individual to improve latent search accuracy, including tenprint, RISC, supplemental fingerprints, and palm prints under the National Palm Print System (NPPS)
  • Rap Back service – regularly monitors and reports on the activities of people under government investigation or supervision, or holding a position of trust such as daycare workers and schoolteachers to authorized organizations to avoid having to repeat background checks
  • Cold Case/Unknown Deceased – enables latent examiners and law enforcement investigators to use the NGI database and search algorithms to identify unknown deceased people in cold cases, such as this 1995 case
  • Iris Pilot – launched in late 2013 to assess its effectiveness, the idea is to use the human iris as biometric identification in correctional facilities, border security, supervised release programs, onsite mobile identification, and investigations involving video evidence; the difficulty is in capturing and compressing the image within acceptable limits

DNA Matching

Genomics revolutionized many things, including criminal investigation, by establishing the idea of DNA profiling. Just like fingerprints, no two DNA profiles are the same. However, it was much more difficult to guard against leaving DNA behind in a crime scene than it was to leave fingerprints behind. Hair, semen, blood, and sweat could contain enough biological material to generate a DNA profile, especially with more advanced technology.

The FBI saw the value of DNA profiles early on in criminal investigation, and eventually built a database called Combined DNA Index System (CODIS). Before there were machine learning algorithms, law enforcement would send trace evidence to special labs to process it, and then submit the DNA profile to the FBI to make a match.

The whole process was a slow one, taking days or weeks, depending on the backlog of the forensics lab, and the difficulty of finding a match. Labs capable of generating DNA profiles were few, and the backlog of cases could be as much as five years in some areas.

Additionally, DNA profiles contain a complex set of data, and CODIS has almost 14 million profiles as of February 2019, so it takes time for standard computer programs to wade through all the data to identify one profile, if it is there. In many cases, the long wait could be justice delayed, justice denied.

A new machine could change all that. Law enforcement has been using fully automated and portable Rapid DNA machines to speed up the process of generating DNA profiles from cheek swabs, cutting it down to just 90 minutes instead of days or weeks. With machine learning software, police could take specimens, generate a profile, and make a match within two hours provided they had access to CODIS.

That is precisely what the FBI means to do by authority of the Rapid DNA Act of 2017. Among other things, this allows the agency to establish a network of these machines to gain access to CODIS.

Here is a 1-minute video showing how the Rapid DNA machine works.

Different Rapid DNA machines came under assessment by the National Institute of Standards and Technology (NIST) in 2018.  Results showed that while the machines are capable of running on its own, a modified analysis with some human intervention yielded results with higher accuracy.


The FBI is the repository of much data spread out over many field offices, so like any big organization, cybersecurity is a major issue. A recent hack by a researcher for a security company revealed just how easy it was to access terabytes of data from the FBI files. The breach allowed access to a multitude of confidential files, from agent interviews to banking transactions.

This is ironic, as the FBI has an entire division dedicated to investigating cybercrime, from ransomware to identity theft. The FBI started contracting cybersecurity company ECS to manage its network security using AI.

AI is particularly indicated when it comes to cybersecurity, primarily because of the rapid evolution of malware. Deep Instinct CTO and co-founder Eli David claimed in an Emerj interview “the conservative estimate is about one million new malwares every single day and it is probably much more than that.” Security software has to be able to keep up with changing parameters, and that is the core of AI.

The $38 million deal would involve six years of ECS will be using machine learning to enable teams of security experts to go through the FBI’s networks to detect weak spots, scan for potential problem areas, and set up staff to assess and manage the security needs of the different networks continuously. ECS would specifically manage the Cybersecurity Red and Blue Team (REBL) and Enterprise Compliance and Continuous Monitoring Support programs.

Insider Threat Identification

Threats from without are bad, but so are threats from within, and maybe much more so. The FBI has been subject to numerous instances of insider threats since the 1980s, a period former Naval Criminal Investigative Service Chief Psychologist and Deloitte Managing Director Michael Gelles calls the “Decade of the Spy.” He described the method of dealing with insider threats as “reactive.”

Gelles further stated:

With the advent of technology and the movement of information, if you will, and the infusion of a new generation into the workforce, information is perceived, managed and shared in a very different way than it was in a world of bricks and mortar — where things now are just instantaneous.

The FBI addressed this need for a more proactive approach to insider threats by launching two programs using data analytics to deal with the problem. One is Javelin, which keeps tabs on internal misconduct, security violations, and internal espionage. The other one is the Insider Threat Analysis Platform (InTAP), which combs through large volumes of data to find patterns that indicate suspicious activities and potential threats to the organization.

Both these programs are underway, but the FBI is still actively trying to recruit AI talent to handle the big data side of them. The problem is competing for the right talent with private companies that can pay more and provide better benefits.

Business Process Management

The FBI is a large organization of multiple field offices across the globe with a staff of more than 51,000 people working around the clock. The old ways of managing the business side of things are bound to fall short with the increasing demands o the times. To reboot its operational efficiency, the FBI has awarded a contract to software company Pega Systems.

Pega would help the FBI streamline operations and cut costs with the use of two of its customizable business solutions: Pega Government Platform and Pega Robotic Process Automation. The platform would be used to develop scalable and flexible apps to meet business process needs. The automation component is to free human staff from repetitive tasks through automation.

According to Pegasystem’s Global Government Business Line Leader Doug Averill:

By overhauling their enterprise business processes with Pega, the FBI can improve the experience for employees so they can spend more time on critical missions that help keep US citizens safe.

The FBI has a long and interesting history, over which it has evolved quite extensively in terms of technological capabilities. AI provides an excellent opportunity for the bureau to maximize the potential of the available data in its various repositories and databases.


Header Image Credit:

Stay Ahead of the AI Curve

Discover the critical AI trends and applications that separate winners from losers in the future of business.

Sign up for the 'AI Advantage' newsletter: