DARPA Seeks AI Solutions to Protect Digital Markets from Potential Exploits

US defense and security agencies, along with federal regulators, are tasked with addressing hacks and data theft, as well as tackling bad actors exploiting gaps within key systems like electronic markets. To counter these threats, DARPA is exploring solutions to identify hidden flaws in these systems.

The Defense Advanced Research Projects Agency has initiated the Intrinsic Market Resilience program. This initiative aims to safeguard electronic markets associated with stocks, bonds, energy, and cloud-computing resources. While these systems have gained speed and efficiency, their intricate connections have introduced new paths for potential disruption.

“I think that the incredible advances in automated reasoning and in AI are opening new ways to defend things,” said Matthew Wilding, the IMR program manager at DARPA. “The focus of this program is using them to defend things, to find those counterexamples and find those improvements we can make to these designs.”

The concern lies not only in potential hacking or data theft but also in exploiting design weaknesses within the markets. Historical instances of exploitation highlight this concern. In the early 2000s, flaws in California’s deregulated energy market contributed to a crisis costing between $40 billion and $45 billion. This incident underscored how manipulation could affect military readiness.

Another example is the 2010 “flash crash,” where US stocks nosedived in minutes due to market instability. While many problems have been fixed, the event demonstrates the impact of weak market designs. The incident involved manipulative trading designs leveraging market quirks.

With more electronic markets and participants, risks of disruptions increase. Additionally, advanced technologies, including AI, have lowered barriers for those with harmful intentions. IMR aims to utilize AI models to study markets, evaluate protections, and identify vulnerabilities.

Importantly, the program targets broader disruption goals, not just individual financial gains. The focus is on bad actors causing economic damage by exploiting regulatory and legal deficiencies. DARPA encourages participants to develop large language models for transforming regulatory documents and technical descriptions into operational computer market models.

The program’s goal is to highlight and correct design vulnerabilities in domestic online equity markets. DARPA wants AI models to propose specific examples of potential market failures, enabling corrections and resilience improvements.

“We’re not just testing,” said Wilding. “We’re actually looking at it and analyzing it, showing how it is resilient, how it is not resilient, and how we could make it resilient.”

This program emphasizes a burgeoning national security concern: future attacks might exploit digital infrastructure, supply chains, or markets, rather than conventional military aggression.