All eyes were fixed on the screen, set to announce the champion. Then, when the Team Atlanta logo appeared, the silence erupted into thunderous cheers and applause. It was the culmination of two years of relentless effort and determination, which had resulted in the team rising above world-class competitors to win the first-place prize of $4 million.

▲ Team Atlanta wins first place at the AI Cyber Challenge.

To take readers behind the scenes of their inspiring journey, Samsung Newsroom sat down with Taesoo Kim, Corporate Vice President, and Joonun Jang from Samsung Research’s Security & Privacy Team, along with Yunjae Choi from the AI Productivity Team.

▲ (From left) Taesoo Kim, Joonun Jang and Yunjae Choi from Samsung Research

Team Atlanta: A 40-Strong Global Alliance of Top Researchers, Engineers and Hackers

The AI Cyber Challenge is a global security technology competition hosted by the U.S. Defense Advanced Research Projects Agency (DARPA). Sponsored by leading tech companies including Google, Microsoft and OpenAI, the latest edition of the event featured a total prize pool of $22.5 million. At this event, Team Atlanta — a coalition of more than 40 world-class security researchers from Samsung Research, the Georgia Institute of Technology, KAIST, POSTECH and other leading institutions — came together under the leadership of Taesoo Kim, who also played a pivotal role in assembling the team.

▲ Taesoo Kim explains what inspired the team to take on the challenge.

Also a professor at the Georgia Institute of Technology, Kim reached out to several of his former Ph.D. students after first learning about the AI Cyber Challenge and encouraged them to join. They soon became the core members of Team Atlanta.

Under Kim’s leadership, Team Atlanta brought together talent from across Korea and both the East and West Coasts of the United States. “Because our members came from such diverse locations, we structured our collaboration framework around each member’s base,” he recalled. “Some of our colleagues in Korea even carved out time from their busy schedules to travel to the Georgia Institute of Technology so we could work side by side.”

How Team Atlanta Outpaced the Competition

The AI Cyber Challenge is a competition designed to test the ability to identify and remediate vulnerabilities in large-scale software systems. Its ultimate aim is to accelerate the development of AI-powered security technologies capable of protecting critical infrastructure such as transportation, energy and healthcare. In the final round, participants were tasked with building an AI-driven “Cyber Reasoning System (CRS)” capable of autonomously detecting software vulnerabilities and generating security patches.

Scoring was mainly based on two components: vulnerability discovery and program repair. Teams earned discovery points for accurately identifying weaknesses in the given code, and repair points for successfully patching them. In short, a team’s score reflected how quickly its system could detect vulnerabilities in the organizer’s source code, how many it could detect and how accurately it could fix them.

▲ The first-place trophy and Team Atlanta’s winning scorecard

Team Atlanta clinched the championship with a decisive lead, uncovering more vulnerabilities and patching them with better accuracy than their competitors.

Jang, who oversaw the development of a large language model (LLM) agent for detecting vulnerabilities in Java programs, explained that the agent helped Team Atlanta’s system detect more vulnerabilities during the competition. “Team members each developed their agents in various ways, and by bringing them together, we were able to build a system capable of covering a wide range of weaknesses,” he said.

▲ Joonun Jang explains the LLM agent Team Atlanta developed for detecting vulnerabilities in Java programs.

Choi, meanwhile, was in charge of developing AI agents that automatically patch vulnerabilities once they are identified. “Since AI was a common tool provided to all participants, true technological excellence lay in how effectively we could harness it,” he noted. “The key challenge was creating AI agents flexible enough to respond to newly discovered vulnerabilities.”

▲ Yunjae Choi explains the AI agents Team Atlanta developed to automatically patch vulnerabilities.

When Seconds Feel Like Eternity

On August 8, the seven finalists gathered in Las Vegas for the final announcement — the culmination of two years of effort. As the world’s largest cybersecurity competition, the event brought together top teams from around the globe, all waiting in tense anticipation. When the third- and second-place winners were revealed, Team Atlanta’s name had still not been called.

“Given how many outstanding teams were competing, I thought that even placing in the top three would have been a great achievement,” said Jang. “So in those moments before the first-place announcement, I was worried but cautiously hopeful at the same time.”

When Team Atlanta’s logo finally appeared on the screen confirming their victory, the audience erupted. “We had only dared to hope, but when we were actually named first place, it was as if the whole team had suddenly found ourselves in the middle of a festival,” said Choi.

“The two years of preparation felt like the shortest time of my life, while the one hour spent waiting for the results felt like the longest,” Kim reflected.

▲ The moment Team Atlanta was announced as the winner of the 2025 AI Cyber Challenge

“I suddenly became emotional when I remembered how, just before the final submission deadline, we had discovered a critical bug in the system we had submitted early to stay ahead of schedule. We worked frantically until the very last minute. I even had to wake the entire team up at 5 a.m. to fix the problem. Just thinking about it still gives me chills,” he added.

▲ (From left) Taesoo Kim, Yunjae Choi and Joonun Jang,

Samsung’s Vision for the Future of AI Security

With its victory at the AI Cyber Challenge, Samsung has once again underscored its technological leadership and global competitiveness in AI security. Going forward, the company plans to further advance next-generation AI security solutions so that they become increasingly capable of autonomously identifying and addressing vulnerabilities.

“Samsung’s AI security technologies are already being applied to strengthen the protection of products and services, including Galaxy devices and AI TVs,” said Kim. “Through continued open-source contributions and active community engagement, we aim to solidify our position as a global leader in AI security.”

The AI Cyber Challenge is a prestigious competition hosted by the Defense Advanced Research Projects Agency (DARPA), a research and development institute of the U.S. Department of Defense, with a total prize pool of $22.5 million.

The finals took place during DEF CON 33, a major hacking and security conference. Universities, research institutions and global tech companies from various countries participated, engaging in a fierce AI security technology competition.

Samsung participated in the competition as Team Atlanta, a joint team led by Samsung Research, the Device eXperience (DX) Division’s advanced research institute, in collaboration with security experts from leading global universities, including the Georgia Institute of Technology (Georgia Tech), Korea Advanced Institute of Science and Technology (KAIST) and Pohang University of Science and Technology (POSTECH).

The competition, which spanned two years starting from 2023, saw forty-two teams compete in the semifinals last year, with only seven advancing to this year’s finals.

“Samsung Electronics was able to achieve meaningful results by demonstrating its AI-based security technology capabilities at a global security technology competition,” said Taesoo Kim, VP of Samsung Research, who led Team Atlanta. “We will continue to expand collaboration with global security experts to further advance Samsung Electronics’ security technology proficiency.”

Team Atlanta advanced to the finals with exceptional technical capabilities, being the sole participant to detect an unintended vulnerability — not a flaw that had been set up for the competition — in the semifinals, and once again proved world-class security technology capabilities in the finals.

The finals evaluated each team’s ability to analyze source code vulnerabilities and apply security patches using AI-powered security technologies alone, without any human intervention.

By showcasing leading expertise with precise and sophisticated AI security technology, Team Atlanta earned the honor of first place and received $4 million in prize.

Building on its achievement, Samsung plans to develop next-generation security solutions powered by AI that can autonomously identify and mitigate security vulnerabilities, enabling continuous enhancement of products and service security.

Samsung formed Team Atlanta — comprised of more than 30 security researchers from the Georgia Institute of Technology, New York University (NYU), Korea Advanced Institute of Science and Technology (KAIST), Pohang University of Science and Technology (POSTECH) as well as Samsung Research, a leading research and development organization within Samsung.

▲ Team Atlanta members pose for a photo.

AIxCC participants compete to detect and mitigate vulnerabilities in large-scale software systems using AI. With a total prize pool of over $20 million, the competition aims to promote the development of new technologies to improve computer code security for social infrastructure systems including transportation, public utilities and healthcare.

The competition is held in conjunction with DEF CON, an annual cybersecurity and hacking conference in Las Vegas. Following the announcement of AIxCC in August 2023, Team Atlanta developed and submitted their cyber reasoning system — before advancing through the semifinals at DEF CON 32 from August 9 to 11, 2024. Securing $2 million in research funds, Team Atlanta will attend the final round at DEF CON 33 in August 2025.

During the semifinal round, Team Atlanta demonstrated its ability to automatically detect security vulnerabilities in large-scale software and apply security patches using their system. Team Atlanta earned significant recognition from organizers as it was the only team out of 39 to discover a genuine vulnerability — not a flaw that had been set up for the competition.

“We will thoroughly prepare for the upcoming final round by diversifying applications of AI and researching programming languages and countermeasures against vulnerabilities,” said Taesoo Kim, Vice President of Samsung Research, who led Team Atlanta.

▲ AIxCC finalists

As cyberattacks become more frequent and complex, the importance of AI technology in security is becoming increasingly evident. By learning from vast amounts of data, AI can analyze and infer from new data to anticipate and respond to security threats. Generative AI, in particular, improves efficiency in the security sector by better understanding software programs and more effectively addressing security threats. As such, the application of AI in security has become a necessity.

Samsung will continue leveraging AI to improve security. Through ongoing research on AI technology that detects vulnerabilities and automates security patches, the company plans to strengthen its competitive edge by increasing product and service reliability.