DARPA wants AI to attack market rules before attackers do. Crypto should pay attention.
DARPA's Intrinsic Market Resilience program funds tools that use large language models and automated reasoning to turn market rules and filings into formal models, search for ways those rules could be exploited, and recommend repairs. It covers markets for stocks, bonds, energy, and cloud-computing resources.
The target is different from the usual one. It isn't rogue profit-seeking manipulation that regulators typically chase. It's actors whose goal is disruption itself. DARPA points to the 2010 flash crash as an example of a design gap with outsized effects. If a model can't prove a market meets a resilience standard, it should generate a specific failure scenario.
Crypto isn't named in DARPA's materials, and the initial focus is domestic online equity markets. The crypto link is an inference. Still, the problem class is familiar. Oracle manipulation, liquidation cascades, and fee or matching quirks are rule exploits, not code hacks.
What it could mean: provable resilience could become a selling point. Venues that can show their liquidation logic and market design hold up under adversarial testing may earn trust, while opaque designs may face more scrutiny.
Crypto also has an unusual advantage. Its rules live on-chain, where anyone, including AI, can read them. That makes protocols easier to model than closed exchange systems, but also easier to attack.
The risk is cost and distance. Formal verification is hard, and this effort is aimed at US equities first. The supportive case is that DeFi teams adopt similar testing voluntarily, before a regulator asks.
In markets, the rules are the attack surface.
$BTC #BTC Price Analysis# $ETH #Altcoin Season#
DARPA's Intrinsic Market Resilience program funds tools that use large language models and automated reasoning to turn market rules and filings into formal models, search for ways those rules could be exploited, and recommend repairs. It covers markets for stocks, bonds, energy, and cloud-computing resources.
The target is different from the usual one. It isn't rogue profit-seeking manipulation that regulators typically chase. It's actors whose goal is disruption itself. DARPA points to the 2010 flash crash as an example of a design gap with outsized effects. If a model can't prove a market meets a resilience standard, it should generate a specific failure scenario.
Crypto isn't named in DARPA's materials, and the initial focus is domestic online equity markets. The crypto link is an inference. Still, the problem class is familiar. Oracle manipulation, liquidation cascades, and fee or matching quirks are rule exploits, not code hacks.
What it could mean: provable resilience could become a selling point. Venues that can show their liquidation logic and market design hold up under adversarial testing may earn trust, while opaque designs may face more scrutiny.
Crypto also has an unusual advantage. Its rules live on-chain, where anyone, including AI, can read them. That makes protocols easier to model than closed exchange systems, but also easier to attack.
The risk is cost and distance. Formal verification is hard, and this effort is aimed at US equities first. The supportive case is that DeFi teams adopt similar testing voluntarily, before a regulator asks.
In markets, the rules are the attack surface.
$BTC #BTC Price Analysis# $ETH #Altcoin Season#
