[Digital Today reporter Chi-gyu Hwang] "Only solidarity beyond borders can stop AI security threats."

Naver Cloud CEO Kim Yoo-won (김유원) said in a keynote speech at Cyber Summit Korea (CSK) 2026 held at COEX in Seoul on Wednesday that security threats caused by AI can no longer be stopped by a single organisation. He stressed that cooperation is needed at the global level.

He said AI has fundamentally changed the nature of security threats. He also made clear that the era of responding to AI attacks at human speed is ending.

Kim said it used to take from tens of days to several months between the discovery of a vulnerability and an actual attack, but now AI finds vulnerabilities and creates attack code on its own. The time required for the process has been reduced to tens of minutes or a few hours.

Deepfakes, sophisticated phishing, automatically generated malware and AI agents that find vulnerabilities on their own and attack have emerged. Keeping AI itself safe, such as against prompt injection, model jailbreaks and data and model contamination, has also emerged as a major security task.

Kim said no single organisation can respond to such threats alone and shared various cooperation cases already under way.

Experts from dozens of countries take part every year in drills hosted by the NATO Cooperative Cyber Defence Centre of Excellence, defending together in real time against large-scale cyber crisis situations. Naver Cloud took part in the drills for a second consecutive year as a member of the South Korean joint team.

In March 2026, the United Nations unanimously launched a standing consultative body to link discussions of responsible state behaviour norms on cyber attacks to actual cooperation and implementation.

He introduced the Defcon hacking contest carried out in the private sector. "At Defcon, hackers from around the world gather to disclose new attack methods and learn from one another. This year, a category that freely uses AI and even a contest in which AI agents directly analyse problems and respond have appeared. Naver Cloud also participated in the contest as a defence team," he said.

Various cooperation ecosystems are also spreading in the private sector. Kim said that a security consultative body formed under Nvidia's leadership in July 2026 includes more than 120 companies, including Naver Cloud, that share AI security technologies and vulnerabilities.

Naver Cloud is also participating in a security-focused AI foundation model project being carried out at the South Korean government level.

Kim said private companies support AI and security capabilities, while the government provides strategic resources such as GPUs. He said the goal is to create models that learn South Korea's unique security environment and industrial field context that general-purpose AI finds hard to understand. He stressed that industries where security is directly linked to national competitiveness, such as semiconductors, finance, power, telecommunications, the defence industry and aerospace, need security models that deeply understand field context.

Kim stressed that the security-focused foundation model will be expanded beyond South Korea to the global market. He said Naver Cloud recently declared the expansion of a global AI factory in cooperation with Nvidia. He said it will make AI security models run at AI factories around the world and connect each base into a cyber security cooperation network. "It is about creating a structure in which multiple countries protect one another rather than one country protecting itself. This is solidarity beyond borders," he said.