AI security specialist AIM Intelligence (CEO Sang-yoon Yoo) announced on the 3rd that it officially participated in the '2025 AI Cyber Defense Contest (2025 ACDC)', the largest AI security event in Korea, and served as a problem setter and seminar moderator for the AI hacking defense contest.
Hosted by the Ministry of Science and ICT and organized by the Korea Internet & Security Agency (KISA) and the Korea Information Security Industry Association (KISIA), this event, with the slogan of 'AI for All, Security for All,' consisted of various programs such as ▲AI Security Insight Seminar ▲AI Hacking Defense Contest Finals ▲AI Security Technology Exhibition.
Aim Intelligence was particularly responsible for creating questions and reviewing the technical aspects of the contest's core programs, "Security for AI" and "AI Hacking Battle." This year's questions expanded beyond text-based LLM vulnerabilities to encompass real-world threat scenarios that could arise in multimodal, agent-based AI systems.
The highlight of the competition, the "AI Hacking Battle," involved teams building their own guardrails and AI systems, then attacking and defending the opposing team's systems in real time. This allowed participants to quickly experience the attack-defense competition of a real-world industrial environment and test their response capabilities.
The presented problems consisted of an attack that causes AI drones to malfunction through visual adversarial inputs, designing a honeypot for an agent, developing an agent that performs an automatic multi-turn jailbreak attack, a scenario that induces an autonomous system to steer in the wrong direction, a data exfiltration attack that steals private keys, and an attack that extracts sensitive information by forcing the call of limited internal tools.
An Aim Intelligence representative explained, “We designed the problem to reflect the risks that can actually occur in multimodal, agent-based AI services.”
Meanwhile, on the first day of the event, CEO Yoo Sang-yoon participated as the official moderator of the 'AI Security Insight Seminar', introducing security issues and the latest trends that arise during the AI adoption process of global companies and organizations, and leading discussions with the speakers.
Aim Intelligence is conducting AI security standardization research and technology collaboration with the Artificial Intelligence Security Institute (AISI), the Financial Security Institute, and the Korea Credit Information Services, and is also strengthening global cooperation as an OpenAI Local AI Safety partner.
CEO Yoo Sang-yoon said, “This is very meaningful as it can contribute to strengthening the AI security ecosystem at the national level,” and added, “We will accelerate the development of technologies to respond to threats that expand to the multimodal and autonomous agent areas.”
- See more related articles
You must be logged in to post a comment.