In an age where technology permeates every aspect of our lives, the promise of predictive artificial intelligence (AI) has captured the imagination of policymakers, law enforcement, and healthcare professionals alike. From self-driving cars to predictive analytics, AI is hailed as the next frontier in solving complex problems. But when it comes to America’s opioid epidemic, a crisis that claims more than 81,000 lives a year, AI has become a dangerous gamble, and the stakes couldn’t be higher.  The U.S. Department of Justice (DOJ), led by figures like Nicole Argentieri, is now leaning heavily on AI-driven solutions to fight the epidemic, wielding complex predictive data systems originally inspired by Wall Street risk management models. The hope is that these systems will identify and disrupt illegal opioid distribution networks, predict where crises will occur, and lead to “maximum harm reduction.” But beneath the technological polish lies a troubling truth.  DOJ’s systems are untested, and in many cases, ill-suited to handle the nuanced, human-centric reality of a national public health crisis.  The DOJ’s aggressive use of artificial intelligence in combating the opioid epidemic stems from several programs designed to track, predict, and intervene in drug trafficking networks. These include the DEA Analysis and Response Tracking System (DARTS), the De-confliction and Information Coordination Effort (DICE), and a variety of federal healthcare task forces utilizing AI systems like NBI MEDIC, Qlarant’s Artificial Intelligence, and CMS Predictive Learning Analytics Tracking Outcomes (PLATO).  The underlying principle is simple, through data fusion, AI systems will flag suspicious patterns of opioid distribution in real-time, enabling law enforcement and public health officials to act swiftly. These AI programs aim to share de-identified data between public health and law enforcement agencies, essentially turning numbers and AI algorithms into actionable intelligence.

But therein lies the rub, because DOJ’s untested systems, inspired by healthcare portfolio insurance models and high-frequency trading AI algorithms, are being repurposed to address a devastating human health crisis.  While AI might be effective at predicting stock market crashes or routing financial investments, it lacks the subtlety and sensitivity required for navigating the opioid crisis, where the consequences of missteps aren’t just monetary, they are human lives.  One of the more troubling aspects of the DOJ’s AI initiative is the influence of Long-Term Capital Management (LTCM)-style risk models. LTCM, a hedge fund that collapsed spectacularly in 1998, relied on similar AI algorithms to manage complex financial risks, only to find that real-world volatility didn’t behave as the models predicted. LTCM’s fall was a warning that no matter how sophisticated a system, models based on historical data and assumptions about human behavior are vulnerable to failure.  The 1987 stock market crash, portfolio insurance, Long-Term Capital Management (LTCM), and the Black-Scholes equation are all tied together by the use of complex AI financial models to manage risk and guide trading strategies. Each of these elements played a critical role in understanding how financial markets behave, especially during times of volatility.  Portfolio insurance was a financial strategy designed to limit losses in a stock portfolio while maintaining potential gains. It became particularly popular in the 1980’s as a form of risk management, especially among institutional investors like pension funds and mutual funds. The strategy typically involves dynamic hedging through the use of derivatives, such as stock index futures and options, to protect against declines in the stock market.

The role of portfolio insurance became a key area of debate in the aftermath. While it was intended to mitigate risk, its mechanical selling triggered massive volatility, highlighting the dangers of relying too heavily on automated strategies in volatile markets.  As discussed, portfolio insurance involved using dynamic hedging strategies to protect portfolios from large losses. These strategies often relied on derivatives, particularly options and futures. The mechanics of portfolio insurance were built on theoretical models like the Black-Scholes formula, which provides a mathematical framework for pricing options. The Black-Scholes model, developed by Fischer Black, Myron Scholes, and Robert Merton in the early 1970s, revolutionized the pricing of options. It provided a formula to calculate the fair value of options, taking into account factors like the stock price, strike price, time to expiration, interest rates, and volatility.  The model assumed that markets are efficient, that prices follow a lognormal distribution and evolve smoothly without jumps, that volatility is constant over time and that there are no transaction costs or liquidity issues.

The 1987 crash demonstrated the limits of these models, especially in extreme market conditions. Portfolio insurance’s reliance on automated selling created a feedback loop that exacerbated the crash. The events of 1987 exposed the flaws in using such strategies without accounting for liquidity risks and the impact of crowd behavior (many players acting in the same way at the same time).  Now, the DOJ is using AI and data fusion, systems birthed from this very model of predictive analytics, in their war against opioids. These programs assume that opioid traffickers will behave predictably, that data patterns will lead to reliable insights into illegal activity. But just as markets can swing wildly due to unforeseen events, so too can opioid networks adapt, evolve, and evade these AI predictive models.  As investors learned from LTCM, algorithms can fail dramatically. When applied to opioid trafficking, a miscalculation could result in missed opportunities to save lives or, worse, the wrongful prosecution of innocent healthcare providers. Imagine the fallout if an AI system, designed to flag suspicious prescribing patterns, inaccurately targets legitimate physicians managing chronic pain patients. The consequences for these individuals, already under intense scrutiny, could be devastating.  One of the cornerstones of these AI systems is their reliance on de-identified data to ensure privacy and confidentiality. The idea is that public health and law enforcement can collaborate without compromising individual privacy by sharing de-identified information. But de-identified data comes with its own set of problems.  AI systems are only as good as the data they process. When data is stripped of its identifiers, critical context can be lost.  AI algorithms might flag certain patterns as suspicious without considering the broader context, such as socioeconomic factors, legitimate pain management needs, or geographic healthcare disparities. De-identified data turns humans into numbers, ignoring the complex realities of the communities suffering from the opioid crisis.  What’s more, the DOJ’s systems over-reliance on such data may push public health decisions into the realm of law enforcement, where every flagged prescription could be seen as a criminal act rather than a cry for help. This blurs the lines between public safety and public health, potentially criminalizing pain patients who are simply trying to manage their conditions.

The opioid crisis is not a problem that can be solved with a one-size-fits-all solution. It requires the delicate balance of human empathy, medical expertise, and data-driven insights. Unfortunately, the current rush to deploy AI systems by the DOJ overlooks the complexities of the opioid epidemic. While the DOJ proudly touts its $1.3 billion black-market drug busts, the real costs are hidden in the stories of patients who are denied care, communities that are over-policed, and lives that are shattered by an over-reliance on technology.  Instead of focusing on harm reduction, these AI systems threaten to deepen the divide between public health and law enforcement, making it harder for those who are already vulnerable to get the help they need. The unintended consequences of using untested AI systems in this context could lead to a public health catastrophe in its own right.  The opioid crisis is a deeply human tragedy. It deserves a response that prioritizes the health and well-being of individuals over the cold calculations of AI systems. While technology can play a role in tracking and understanding patterns of drug trafficking, it must be carefully tested, regulated, and balanced by human oversight. We cannot allow the same reckless reliance on predictive models that led to financial ruin in the past to guide our response to a crisis of this magnitude.  The DOJ’s embrace of AI as a silver bullet is a dangerous miscalculation. Instead, we must demand transparency, ethical oversight, and above all humanity in our approach to solving the opioid epidemic. Because when lives are at stake, we cannot afford to leave our fate in the hands of untested machines.