The Pentagon’s Defense Advanced Research Projects Agency is turning to artificial intelligence to fortify the electronic systems that underpin everything from stock trading to energy distribution. DARPA’s new AI market resilience program, called Intrinsic Market Resilience, seeks tools that can spot hidden structural weaknesses in digital markets and propose fixes before adversaries can weaponize them for economic havoc.
From Market Efficiency to Market Fragility
Electronic markets have grown faster and more interconnected, DARPA notes, but that same connectivity has opened new paths for disruption. The concern is not primarily hacking or data theft—threats that have long been on the radar—but rather the exploitation of flaws baked into the design of the markets themselves.
Matthew Wilding, who manages the Intrinsic Market Resilience program at DARPA, said advances in automated reasoning and artificial intelligence are creating fresh ways to defend critical systems. He acknowledged that the same technologies can be used offensively, but stressed that the program’s focus is defensive: finding counterexamples and identifying improvements to market designs. The goal, he said, is to make these systems more resilient and dependable.
Lessons From California and the Flash Crash
DARPA points to two episodes in recent decades to illustrate the stakes. In the early 2000s, traders took advantage of weaknesses in California’s newly deregulated energy market, contributing to a crisis that DARPA says cost between $40 billion and $45 billion. The upheaval was severe enough that the federal government invoked the Defense Production Act to secure natural gas supplies for military installations—a reminder that manipulation of market mechanisms can ripple into military readiness.
The second example is the 2010 “flash crash,” when US stock prices plummeted within minutes. Investigators later linked the turmoil in part to manipulative trading by one individual, among other factors. According to DARPA, that person’s sell orders were crafted to avoid execution by exploiting a quirk in market design, so the actions cost the individual nothing. Many of those specific gaps have since been closed, but the episode underscored how design flaws can be gamed.
A Growing Attack Surface
The number of electronic markets and participants continues to climb, raising both the likelihood of disruption and the potential damage from exploitation. At the same time, the spread of AI and other technologies has lowered the barrier to entry for malicious actors. DARPA’s program aims to use some of those same tools to model markets, analyze their protections, and flag weak points.
This is not about the routine market manipulation that regulators typically pursue, such as individuals seeking financial gain. Instead, DARPA is targeting adversaries whose aim is disruption, economic damage, or instability—actors who can leverage gaps in legal and regulatory safeguards.
Modeling Markets With Language Models
Under the program, performers are expected to use large language models to transform regulatory filings and technical specifications into computer models of how markets operate and interact. The agency wants participants to identify design flaws in domestic online equity markets and recommend repairs to protect against those seeking to exploit market design vulnerabilities.
If a model cannot demonstrate that a market meets a certain resilience standard, DARPA wants it to generate a concrete example of how the market could fail. That would let users pinpoint the weakness and propose a fix. Wilding described the effort as more than testing: the program is analyzing how a market is resilient, how it is not, and how it could be made resilient.
The solicitation reflects a broader national security concern that future attacks on the United States may not look like conventional military action. Instead, they could exploit vulnerabilities in digital infrastructure, supply chains, or markets—making the resilience of electronic market design a matter of defense.
