US intelligence community holding a contest to help build software that predicts hacks
Aliya Sternstein / Nextgov :
Context & Ripple Effects
This 2015 report is the earliest data point in a decade-long arc of US government hacking competitions. The intelligence community's contest to predict hacks predates DARPA's $2M Cyber Grand Challenge, which asked teams to build AI that exploits rivals' flaws while patching its own.
The pattern matured in 2023, when the White House backed DARPA's AI Cyber Challenge debut with Anthropic, Google, Microsoft, and OpenAI enlisted to build systems that find and fix open-source flaws — making this early intelligence-community contest the template's origin.
First-order effects
- Contest participants gain a funded path to build predictive-hack software for intelligence-community buyers, converting prize money into a talent and tooling pipeline the agencies do not have to develop in-house.
Second-order effects
- Commercial security vendors face contest-bred tools entering the market as free or government-sponsored alternatives, pressuring the pricing of vulnerability-scanning products.
- The success of a prize model shifts agency procurement from buying finished tools to sponsoring competitions, as DARPA later did at far larger scale.
Third-order effects
- If the competition model holds — and the 2023 AIxCC lineage suggests it does — US cyber capability consolidates around recurring public contests that seed dual-use code intelligence across government, labs, and the open-source ecosystem.
The trend: US agencies are sourcing offensive and defensive cyber AI through recurring public competitions rather than internal development, with each contest scaling up the last.