Why AI Will Make Law Enforcement Fairer and Society Safer

AI is already bringing efficiencies to vast sectors of the economy including finance, insurance, healthcare, customer service, manufacturing, and more. But when it comes to law enforcement, much of the conversation has revolved around potential downsides such as algorithmic bias, error-prone facial recognition technologies, and other ways AI could be used, intentionally or not, to improperly tilt the scales of justice.

While these are real issues that must be addressed, the truth is that AI has the potential to make our criminal justice system much fairer and more efficient, and our communities safer. AI is an ideal tool to address a massive, under-appreciated, and all too real threat to the justice system: the inability of police, prosecutors, defense attorneys, and others to make sense of the staggering amount of evidence that’s available in the modern world. 

Whatever you think about the explosive growth of text, audio, video, location data, and other information coming off our phones, smart doorbells, surveillance cameras, sensors and IoT devices — not to mention, social media posts — it’s real and accelerating daily. AI is the only technology that can be used to quickly distill this vast ocean of disparate data points into a foundation of fact that can serve as the basis for a more accurate and more trusted approach to law enforcement.

“AI is the only technology that can be used to quickly distill this vast ocean of disparate data points into a foundation of fact that can serve as the basis for a more accurate and more trusted approach to law enforcement.”

From where I sit, we ought not let hypothetical misuses of AI stand in the way of the very real benefits it can bring today. We can, and must, develop guardrails and methods to make sure unethical people don’t use the technology to support their own biases or selfish purposes. But when used properly, AI can be instrumental in making sure that guilty criminals are brought to justice, and that erroneously accused defendants are washed-out of investigations as early as possible and not subjected to trial.

From intelligence, to cyber security, to law enforcement

My seat happens to offer a pretty interesting vantage point on this topic. I started my career tracking cyber attackers at the National Security Agency right before 9-11 and ended up as the technical director (essentially, the CTO) of counterterrorism at the agency about 15-years later. I saw first-hand how the best intelligence technologies could be used to piece together complex evidence in near real-time to pursue justice, including the capture of Osama bin Laden.

Two decades and four security-related start-ups later, I started Tranquility AI with my co-founder Dave Harvilicz to put amazing new AI technologies to work in the civilian law enforcement world. We knew the depressing reality facing police departments and prosecutors. Far too often, they decide which cases to pursue not based solely on the facts but on whether they have the time and resources to gather and analyze all the relevant evidence to establish those facts beyond a reasonable doubt.

For starters, there are not enough good guys. The number of prosecutors in the U.S. has not budged since 2000, and police departments have been dealing with a huge labor shortage due, in part, to a wave of early retirements and a decline in recruits that began in the aftermath of the protests against police violence following the death of George Floyd in 2020.

At the same time, dealing with the tsunami of data generated by our digital world has become overwhelming. For example, when detectives get a warrant to search a suspect’s cell phone, they typically get back a 250,000 page PDF file. The legacy analysis tools used by these professionals to mine such digital evidence means they spend hours every day tediously looking for needles in haystacks — needles that sophisticated AI models are perfectly suited to find.

A powerful test-case in New Orleans

We got our first chance to test our thesis that AI could help tackle these problems in 2023, when New Orleans District Attorney Jason Williams brought us in as part of a team of consultants to help deal with a growing backlog of cases due to a COVID-era spike in violent crime. As in many other cities, less than half of violent criminals in New Orleans were being brought to justice. The odds of getting away scot-free were so good that many suspects thought little of brazenly doing financial transactions on Venmo or Zelle, or of dropping hints about their crimes on Instagram. We developed a process to quickly corroborate these and other clues. Rather than spend hours of digital drudgery combing through mountains of data, Williams’ team could get a concise summary of the evidence in a matter of minutes.

“When detectives get a warrant to search a suspect’s cell phone, they typically get back a 250,000 page PDF file.”

Based on the success of that project, we productized that system into a platform called TimePilot that simplifies and improves the process. TimePilot ingests a wide range of data types, and converts it into rich media timelines that provide visualizations of how crimes unfolded, annotated with everything from police body cam footage to social media posts. New Orleans quickly signed on as our first customer, and we’ve received dozens of calls from police and prosecutors since The Wall Street Journal published an article about our work in the Big Easy in late 2023.

Prosecutors have benefited greatly from being able to interrogate their case files using natural language questions and getting answers within seconds that would have taken hours to look up manually. I’ve been told repeatedly that it feels like they are talking to the Star Trek computer when using TimePilot’s chat to prepare for charging, plea bargains, and trials.

The massive opportunity behind a just cause

So far, there’s been no rush of new start-ups into the $20 billion market for law enforcement software, no doubt due to the many difficulties of selling into such a high-profile, regulated, politicized, and cash-strapped industry. For good reason, police departments and prosecutors tend to be conservative when it comes to trying new technology that could potentially be abused. 

We hope to help change this mindset, because the opportunities for technology companies to positively impact the world with AI are tremendous.

1. Fairness 

From the start, our product development focus was on accuracy, to make sure that AI is only used in an impartial way. In fact, our mission statement is “accurate justice for all.” Far from introducing bias, our goal is to provide facts and incontrovertible evidence that humans would be hard-pressed to correlate and discover on their own. In one recent case, an individual was incorrectly named as a suspect in a murder because he had an on-screen handle that was only slightly different from that of a known criminal. TimePilot quickly revealed that the person could not have been involved in the crime, based on a discrepancy in the birth date and mobile phone known to be used by the true perpetrator, and the innocent’s name was removed from the case before it was brought to court. 

One concern people have is that AI will take the reins in future investigations. To be clear, we never want AI to call the shots on who to charge for what crime. We believe justice should always be decided by people – police, prosecutors, juries and investigators. Judges and jurors still need to evaluate any and all evidence unearthed by AI systems like ours. We have no interest in using AI to do predictive “Minority Report” style law enforcement. When used properly, AI will help professionals do their jobs more fairly and accurately.

AI will also make life more difficult for those bent on abusing their power or on letting their own biases or preconceptions influence their work. For example, TimePilot can ensure more consistent compliance with the Brady Rule, which requires that all exculpatory evidence be turned over to defense counsels, by automatically prompting prosecutors to share such information as it is discovered. And we offer the same capabilities we sell to law enforcement to defense lawyers, giving them the tools they need to more easily spot and neutralize unethical or unjust prosecutions. Let’s face it; the justice system is rather opaque. If a prosecutor knows going into trial that the case has an evidentiary weak spot, they don’t share that with defense counsel. By taking advantage of AI, the defense counsel may not need them to.

The ability to gather and synthesize data quickly will also contribute to fairness. Defendants won’t have to sit in jail waiting for prosecutors to decide whether to bring a case if it will ultimately not be prosecuted.

2. Reduced crime

AI systems like TimePilot will result in more crimes being prosecuted and punished. Law enforcement will waste less time on cases that are ultimately dropped, and will be able to make cases far more efficiently. This will matter to criminals, who will know that evidence will not go unseen. In fact, one defendant accepted a 30-year plea deal after seeing how the New Orleans DA used AI to make its case. I believe AI is one reason violent crime fell even faster in that city than in the nation overall as we came out of COVID-era lockdowns.

“In fact, one defendant accepted a 30-year plea deal after seeing how the New Orleans DA used AI to make its case.”

AI will also empower local law enforcement to more effectively address common crimes that routinely fall through the cracks. For example, car theft has become an epidemic in cities such as Denver and San Francisco, but often goes unprosecuted because “the juice is not worth the squeeze”, leaving law-abiding residents and victims understandably furious. The same goes for cyber-crime, which costs American citizens trillions of dollars a year. Police departments often don’t have cyber-specialists on staff, and the FBI typically doesn’t pursue cyber-fraud cases unless the damage exceeds $75,000. Contrary to popular belief, not all cyber-criminals are overseas and local law enforcement can and should investigate these crimes to identify the perpetrators, arrest them, and extradite them back to face justice when they remotely attack unwitting victims. AI means local law enforcement can now pursue these criminals, whether they are sophisticated Russian crime-rings or a cyber fraudster down the street. 

3. Efficiency

AI can increase the bang for the buck that local, state and federal governments get from their investments in law enforcement. As with cyber-crime, it can reduce the cost of criminal activity in monetary terms. And it will transform the productivity of everyone in the justice system. New Orleans DA Williams said our system saved his team thousands of hours of fruitless digital sleuthing over the course of a year. No doubt, they saved countless more hours because they were able to determine far more rapidly which cases to drop. 

Preparation for trial and plea bargaining can take prosecutors a month of wall-clock time to be fully-prepared, sometimes longer based on the complexity of the case. Using TimePilot, the time to negotiate a plea is cut down from 30 days to 3 days on average, leading to higher throughput in the justice system.

4. Renewed faith in the justice system

According to a recent Gallup poll, only 17% of Americans have a great deal or fair amount of confidence in the criminal justice system. That places it among the lowest rated institutions, below newspapers, public schools and large tech companies. Confidence is shaken in institutions when subjectivity overtakes objectivity with respect to decision making. In order for confidence to increase in the criminal justice system, AI can help restore objectivity and “blindness” in the appraisal of crime. Removing the capricious attributes of prosecution and policing will increase confidence in the long-term.

“AI can help restore objectivity and “blindness” in the appraisal of crime. Removing the capricious attributes of prosecution and policing will increase confidence in the long-term.”

If properly implemented, we are confident AI can help improve this sad state of affairs. Criminals, especially those who have come to believe they can routinely beat the system, will know that the odds are no longer in their favor. Citizens who’ve lost faith in the police’s ability to stop common crimes and prosecutors to convict criminals will see their faith renewed. Over time, AI may even help stimulate the conversation about which laws ought to remain on the books and be enforced versus which laws are outmoded and ought to be repealed.

Law-abiding citizens, good police, overworked prosecutors, defense attorneys and innocent defendants, will all benefit. Bad cops and criminals will not.