 |
LAS VEGAS, Aug. 11, 2025 /PRNewswire/ — Offensive security firm Theori has taken third place at the finals of the Defense Advanced Research Projects Agency’s (DARPA) AI Cyber Challenge. The $1.5 million prize was awarded at DEF CON 33 in Las Vegas, one of the world’s most influential cybersecurity conferences.
Theori receives $1.5M prize at DARPA’s AI Cyber Challenge
The AI Cyber Challenge (AIxCC) is a competition led by DARPA and the Advanced Research Projects Agency for Health (ARPA-H). It tasks elite security researchers with creating artificial intelligence that can automatically find and repair security flaws in the world’s most critical open-source software.
Theori’s AI system, “Robo Duck,” earned its top-tier ranking among seven elite finalists from leading academic and commercial institutions. In the final evaluation, the system autonomously identified a total of 34 vulnerabilities in real-world software, operating entirely without human intervention. Notably, this included zero-day vulnerabilities, critical flaws completely unknown to developers and for which no defense exists.
This achievement follows Theori’s dominant first-place victory in the 2024 semifinals, where the team won a $2 million prize from a field of over 100 teams.
“This represents a pivotal moment in securing the software that powers our world,” said Brian Sejoon Pak, CEO of Theori. “By automatically finding zero-day vulnerabilities in this competition, our work proves that the combination of deep offensive security expertise and advanced AI is already delivering critical security results today, addressing complex flaws that were previously impossible to solve at scale.”
Adding to the achievement, Theori also secured victory in the renowned DEF CON Capture the Flag (CTF) competition, marking their ninth total victory and a record-setting fourth win in a row. At DEF CON, the company was also awarded Best Privilege Escalation at the 2025 Pwnie Awards — the “Tonys” of cybersecurity — for its discovery of CVE-2024-50264, a critical Linux kernel vulnerability that could grant attackers root privileges. The team’s repeated dominance in the CTF, along with recognition for impactful vulnerability research, demonstrates a level of human expertise that the company has now proven it can translate into its AI-powered system.
In line with DARPA’s commitment to the community, Theori’s system will be released as open source to improve access to advanced cybersecurity tools. More details on the “Robo Duck” system are available at: https://theori-io.github.io/aixcc-public
About Theori
Theori is a pioneering cybersecurity firm based in Austin, Texas, specializing in offensive security research, red teaming, and AI-driven security solutions. Trusted by over 100 global organizations including Google, Microsoft, and Okta, Theori’s automated security platform enables continuous vulnerability detection at scale, making enterprise-grade security accessible to organizations of all sizes. With a team of world-class hackers, the company provides unparalleled expertise in identifying and addressing complex security challenges to help organizations strengthen their security posture.
