The FBI Is Exploring AI to Identify Future Criminal Threats

The idea sounds like it belongs in a science fiction thriller. A government agency uses artificial intelligence to identify people who could become threats before they break the law. Yet that possibility is moving closer to reality as the FBI explores new AI capabilities for its Threat Screening Center, triggering fresh debates over privacy, surveillance, and the limits of predictive policing.

Supporters argue that AI could help investigators detect dangerous patterns far faster than humans ever could. Critics counter that asking software to predict future criminal behavior risks expanding government surveillance while making life-changing decisions based on probabilities rather than evidence.

The FBI Wants AI to Strengthen Threat Screening

The controversy stems from procurement documents showing that the FBI’s Threat Screening Center is seeking an artificial intelligence platform capable of enhancing its existing operations.

According to the request, one of the core capabilities sought is “Predictive Modeling Using Enhanced Data with Traceable Lineage.” The goal is to analyze patterns across government datasets and identify individuals who may warrant closer examination.

Rather than replacing investigators, the proposed technology would automate work currently performed manually by analysts reviewing vast amounts of information.

The Threat Screening Center already plays a central role in maintaining the federal terrorist watch list, sharing information across agencies responsible for aviation security, border protection, and law enforcement. The addition of predictive AI would represent a significant expansion in how those existing datasets are analyzed.

While predictive technology has become increasingly common in sectors such as finance, healthcare, and cybersecurity, applying similar techniques to national security introduces a very different set of ethical and legal questions.

The Bureau Says AI Is Already Part of Modern Investigations

Although recent reports have focused on predictive policing, the FBI has repeatedly emphasized that artificial intelligence is already used in more limited investigative roles.

According to the bureau, AI helps process enormous volumes of digital evidence that would otherwise require extensive human labor. Current applications include vehicle recognition, language identification from voice recordings, speech-to-text transcription, and video analytics.

The FBI states that these systems are designed to generate investigative leads rather than make investigative decisions.

The agency also stresses that every AI-generated result is reviewed by trained investigators before any substantive action is taken.

Its public guidance states that “A human being is ultimately accountable for the actions taken, not an AI.”

The bureau further explains that all AI-generated investigative leads undergo verification and validation by human experts before they are acted upon.

Officials describe this human oversight as essential because investigations require a very high degree of certainty and because AI systems remain susceptible to errors.

Why the FBI Believes AI Has Become Necessary

The bureau argues that artificial intelligence presents both an opportunity and a growing threat.

According to its public guidance, criminals increasingly use AI to automate cyberattacks, conduct sophisticated fraud, generate convincing deepfakes, and support other criminal activities.

The FBI notes that AI itself is not illegal. Instead, it warns that malicious actors are rapidly adopting inexpensive AI tools capable of producing convincing fake content and automating attacks that previously required substantial technical expertise.

As more businesses and government agencies adopt artificial intelligence, the bureau expects the potential attack surface to expand as well.

Officials say law enforcement must evolve alongside these technological changes rather than respond after criminals have already adopted new capabilities.

The bureau groups its AI strategy into three primary objectives.

The first focuses on identifying criminal and adversarial uses of AI, including cybercrime, fraud, violent offenses, and threats to national security.

The second aims to protect American innovation by preventing foreign actors from stealing AI technologies developed in the United States.

The third centers on responsible internal use of AI to help investigators process large volumes of information while remaining under human supervision and complying with constitutional protections.

A Growing Watch List Could Make Automation Attractive

The renewed attention on predictive AI comes as the federal watch list has grown dramatically.

Reports indicate that the database now contains nearly two million names, making manual analysis increasingly difficult.

Maintaining a system of that scale requires analysts to review enormous quantities of intelligence, travel information, biometric records, and investigative data.

Supporters of AI-assisted analysis argue that software could identify relationships and behavioral patterns much faster than human investigators.

From their perspective, AI would not be deciding guilt or innocence. Instead, it would prioritize information that deserves additional review.

FBI Director Kash Patel has publicly expressed enthusiasm for expanding artificial intelligence across bureau operations.

Speaking about modernization efforts, Patel said, “I’ve got every major tech company in the world embedded in the FBI,” describing efforts to integrate AI into counterterrorism work to deliver “instantaneous results.”

Those comments have fueled speculation about how extensively artificial intelligence may eventually be incorporated into future investigations.

Even so, procurement requests represent an early stage of the acquisition process and do not necessarily mean a specific system will ultimately be purchased or deployed.

The Biggest Concern Is Not the Technology Itself

Much of the criticism surrounding predictive AI centers on the data powering it rather than the algorithms themselves.

Machine learning systems identify patterns by analyzing historical information. If the underlying data contains inaccuracies, incomplete records, or historical biases, those problems can influence the system’s recommendations.

Civil liberties advocates argue that predictive models can unintentionally reinforce existing disparities by relying on data generated through previous policing practices.

Organizations such as the Brennan Center for Justice have warned that predictive data-fusion systems carry heightened risks involving bias, surveillance, and false positives.

Critics also point to previous errors involving federal watch lists.

Government audits have documented mistakes within watch-list databases over the years, while court cases have challenged individuals being placed on those lists without sufficient justification.

Opponents argue that introducing predictive modeling on top of imperfect datasets could increase the likelihood of innocent individuals receiving additional scrutiny.

Supporters counter that these concerns strengthen the case for human oversight rather than eliminating AI entirely.

They argue that technology should assist investigators, not replace their judgment.

Privacy Advocates Say Prediction Is Different From Investigation

The debate surrounding the FBI’s proposal goes beyond artificial intelligence itself. The larger concern is whether government agencies should use predictive systems to assess future risk before an individual has committed a crime.

Critics argue that traditional investigations begin with evidence of criminal activity. Predictive models, by contrast, attempt to estimate the likelihood that someone could become a future threat based on patterns hidden within large datasets.

That distinction has prompted civil liberties organizations to question how such systems might affect constitutional protections.

The Brennan Center for Justice has previously classified predictive data-fusion technologies as high-risk because of concerns surrounding bias and indiscriminate surveillance. According to reports, earlier federal guidance under both the Biden and Trump administrations acknowledged those risks when evaluating AI procurement.

Some observers also note that requirements for bias testing in certain federal AI purchases have since been removed, increasing concern among privacy advocates who believe independent evaluation is necessary before predictive systems become operational.

For those critics, the central issue is accountability.

If an algorithm incorrectly identifies someone as a potential threat, understanding why that decision occurred can become extremely difficult, particularly if the underlying model relies on thousands of variables interacting simultaneously.

The Data Behind the Predictions Is Also Under Scrutiny

Another aspect of the debate focuses on the information feeding these systems.

Reports indicate that the FBI already purchases certain commercially available location data and other information from data brokers, a practice that has attracted attention from lawmakers and privacy advocates alike.

Combined with biometric information, travel records, intelligence reports, and other government databases, AI systems could potentially analyze enormous quantities of information within seconds.

Supporters say that capability could help investigators identify genuine threats much earlier than traditional investigative methods.

Critics see another possibility.

Rather than simply identifying dangerous individuals, they worry that predictive systems could assign risk scores to people who have never committed a crime, expanding surveillance without sufficient evidence.

The distinction matters because prediction is inherently probabilistic. Even highly accurate models produce false positives, meaning innocent people can be incorrectly flagged for additional review.

The larger and more complex the database becomes, the greater the challenge of ensuring every recommendation is accurate.

AI Is Already Transforming Other Parts of the Justice System

The controversy surrounding predictive policing stands in sharp contrast to other criminal justice applications where artificial intelligence has produced measurable public benefits.

One example comes from Rasa Legal, a technology company founded by former public defender Noella Sudbury.

Rather than predicting future crime, Rasa uses AI to help people clear eligible criminal records more efficiently.

Sudbury developed the platform after years of seeing clients struggle to secure employment and housing because of old convictions that could legally be expunged.

The traditional process required lawyers to manually review extensive court records before determining whether someone qualified for record clearance.

Using AI, that review can now be completed in approximately three minutes.

According to Sudbury, automation has reduced the time required to prepare expungement petitions by more than half, allowing legal aid organizations to assist far more people than would otherwise be possible.

Since launching in 2022, the company says it has helped clear the records of more than 34,000 people across Utah, Arizona, and Pennsylvania.

Research cited alongside the project suggests that stable employment and housing reduce the likelihood of future criminal behavior. Studies have also found that wages increase substantially following successful record expungement.

For Sudbury, AI represents an opportunity to remove barriers rather than create new ones.

“What that means for me as a lawyer is that we can scale our impact far beyond what would be possible if we were relying on human beings,” she said.

The contrast highlights a broader reality about artificial intelligence.

The same underlying technology can either expand opportunity or increase scrutiny, depending on how governments, businesses, and institutions choose to deploy it.

The Future May Depend on How Human Oversight Is Applied

Despite headlines suggesting machines could soon determine who becomes a suspect, the FBI maintains that human investigators remain responsible for every substantive decision.

Its public guidance repeatedly emphasizes that AI-generated results are investigative leads, not evidence.

The bureau also states that its policies governing AI are designed to protect privacy, civil liberties, ethical standards, and constitutional rights while ensuring every recommendation receives human review before action is taken.

Those assurances have done little to quiet concerns among critics who argue that once predictive systems become embedded within government agencies, their influence can gradually expand beyond their original purpose.

Technology experts often point out that AI models improve over time as they receive additional data. Whether that evolution strengthens investigations or broadens surveillance is likely to remain a subject of intense political, legal, and public debate.

The FBI’s procurement request does not guarantee that a predictive system will ultimately be deployed. It does, however, signal that federal law enforcement is actively exploring how artificial intelligence could reshape national security operations in the years ahead.

The discussion now extends far beyond one agency or one procurement document. As AI becomes more capable, governments around the world will face increasingly difficult choices about balancing security with individual rights. How those decisions are made could shape public trust in artificial intelligence just as much as the technology itself.