South Korea is joining Anthropic's AI-powered cybersecurity collaboration project amid growing concern over generative AI's expanding role in both cyberattacks and defense. The government and major domestic companies are entering the global AI firm's security network, with AI-assisted vulnerability detection and response systems expected to expand as a result.
Anthropic said Tuesday it is broadening participation in Project Glasswing — which uses its cybersecurity-focused AI model "Claude Mythos Preview" — to more than 150 new organizations across at least 15 countries.
Glasswing is a collaborative program launched in April to identify security flaws in critical software and support patching efforts using AI models.
Initially, about 50 partners received access to the Mythos model to review their codebases. According to Anthropic, the process uncovered more than 10,000 high-severity or critical vulnerabilities. The company said new organizations must meet separate security requirements before gaining access.
South Korea is among the newly added participants. The Ministry of Science and ICT is joining through the Korea Internet & Security Agency, known as KISA. Among domestic companies, Samsung Electronics, SK Hynix and SK Telecom are included in the expansion. However, the companies' official responses vary. Samsung Electronics said it could not confirm the details, while SK Hynix said nothing had been finalized.
Anthropic said the new cohort includes industries that were relatively underrepresented among early partners — among them power, water, healthcare, telecommunications and hardware. The rationale is that software or infrastructure managed by these organizations, if attacked, could cause damage that spreads across borders, making preemptive defensive cooperation necessary.
The rapid advancement of AI security models is driving Anthropic's push to accelerate Glasswing's expansion. The company warned that other AI firms could have Mythos-level models within the next six to 12 months, and that if similar models are released without adequate safeguards, cyberattacks could become more frequent and harder to predict.
AI-based cybersecurity collaboration is also broadening within South Korea. As the government and major companies increasingly join programs that leverage global AI firms' security models, the emphasis is shifting from personnel-driven security responses toward systems that combine AI-based vulnerability detection and patching.
rim@heraldcorp.com
