SKT, Upstage Launch 'Sovereign AI Cybersecurity Council' with Security Firms, Universities

SK Shieldus, AhnLab, Korea University and 8 other organizations participate
"AI models to be trained on field data from security companies"

Participants pose for a commemorative photo at the launch ceremony of the Independent AI Cybersecurity Council, held at SKT Tower on the afternoon of the 7th. From left: Choi Yongjin, head of the AI Tech Center at SK Telecom; Choi Donghee, head of the AI Strategy Planning Office at SK Telecom; Choi Daeseon, dean of the College of AI at Soongsil University; Heo Sinwook, professor at Pusan National University; Kim Taeyoon, who oversees foundation models at SK Telecom; Lee Jaeyong, head of the public sector division at Upstage; Yoo Kyungsang, head of the AI CIC at SK Telecom; Jung Ikrae, dean of the Graduate School of Information Security at Korea University; Kwon Heonyoung, professor at Korea University; Sukkyoon Kang, CEO of AhnLab; Jung Samyong, CEO of Secui; Cho Youngchul, CEO of PIOLINK; Lee Dongbeom, CEO of Genian; Lee Jeonga, CEO of RaonSecure; and Kim Byungmoo, vice president of SK Shieldus. Provided by SKT.

SK Telecom said on October 8 that it had launched the Sovereign AI Cybersecurity Council with Upstage, major information security companies in Korea and universities, and signed a business agreement to cooperate on AI information security technologies.

A total of 11 organizations took part in the agreement ceremony, held at SKT Tower on the afternoon of October 7. They included SK Telecom and Upstage, six information security companies—SK Shieldus, SecuI, AhnLab, RaonSecure, Genian and Piolink—and three universities: Korea University, Soongsil University and Pusan National University.

The agreement aims to extend SK Telecom's A.X K series, which advanced to the third stage of the government's Sovereign AI Foundation Model Project, and Upstage's Solar model into cybersecurity. Under the council's framework, the two models will be trained on field data from security companies, their safety will be verified by universities, and their performance will be assessed through application to participating companies' security products and services.

SK Telecom and Upstage plan to train their AI models on security operations data generated in real-world security settings, including malware analysis, vulnerability analysis and security monitoring, provided by participating organizations. The goal is to improve the AI's ability to understand and assess security situations.

Training will take place in two stages. During pretraining, security data will be used to strengthen the models' foundational knowledge. SK Telecom's next model, A.X K3, will systematically learn from data in the security field, and its training data will also be supplemented to help ensure the safe operation of its AI for Everyone service, which is aimed at the entire population. Upstage will focus on training its next-generation Solar model, which can analyze large volumes of code and long-term logs at once, on security data for vulnerability detection and analysis.

In the subsequent post-training stage, operational data from security environments will be used to improve the models' ability to perform real-world tasks. These include analyzing and patching software vulnerabilities, responding to incidents at security operations centers, and using security tools safely. Participating information security companies will bring their expertise in areas such as threat intelligence, malware analysis data and security solution operations to model development and validation.

Participating companies will apply A.X K and Solar to their security products and services, including zero-day vulnerability analysis and real-time security monitoring, and conduct joint trials in live operating environments. The trials will assess the models' performance and safety, and insights gained from analysis and response in the field will be fed back into training. The council plans to establish a virtuous cycle linking data collection, model training and performance validation.

The council will use "mutual red teaming," in which A.X K and Solar attack and evaluate each other to identify vulnerabilities. Cases uncovered in this process will be used as training data to strengthen both models' defensive capabilities. Participating universities will design and evaluate the red-teaming process as independent third parties. Korea University will assess the AI's ability to analyze executable files and defend against attacks; Soongsil University will assess AI security performance standards and the ability to counter attacks that bypass safety measures (guardrails); and Pusan National University will verify the performance and safety of next-generation and lightweight models.

Yoo Kyungsang, head of SK Telecom's AI CIC, said, "Models validated through the Sovereign AI Foundation Model Project will demonstrate their effectiveness in real-world security settings alongside leading Korean information security companies and universities. We will also show how the country's flagship AI models can be put to use in cybersecurity."

This content was produced with the assistance of AI translation services.