SKT Enlists Gated Claude Mythos AI to Shield Core Infrastructure

Why It Matters

The partnership gives SKT a cutting‑edge defensive tool that could dramatically reduce breach risk for its massive subscriber base, while bolstering the broader ecosystem’s resilience against AI‑enhanced cyber threats.

Key Takeaways

• SK Telecom allocates $453 M over five years for AI‑driven cybersecurity
• SKT joins Anthropic's Project Glasswing, accessing Claude Mythos Preview
• Project Glasswing has uncovered >10,000 critical flaws in its first month
• South Korea's telecoms, Samsung and SK Hynix, strengthen national cyber defenses
• Anthropic pledges $100 M in usage credits and $4 M donations for open‑source security

Pulse Analysis

Anthropic’s Project Glasswing marks a turning point in how AI is deployed for defensive cybersecurity. By granting select partners access to Claude Mythos Preview—a model deemed too powerful for public release—Anthropic enables automated, multi‑step attack simulations that would normally require weeks of expert labor. The initiative’s rapid expansion to 150 new organizations across sectors such as power, healthcare, and communications underscores a growing consensus: AI can accelerate vulnerability discovery, but only if it is harnessed responsibly and under strict governance.

For SK Telecom, the collaboration is more than a technology upgrade; it is a strategic response to a 2025 data breach that exposed 23 million USIM records. The $453 million budget earmarked for AI‑based monitoring and a Zero‑Trust architecture reflects a broader industry shift toward proactive, model‑driven security postures. By integrating Claude Mythos into its security operations center, SKT can continuously scan codebases, prioritize critical flaws, and patch them before malicious actors exploit the same pathways. This approach not only protects its own network but also contributes to the security of downstream services that rely on SKT’s infrastructure.

The ripple effects extend to global supply chains and national security. Anthropic reports that partners have already identified more than 10,000 high‑severity vulnerabilities, many persisting for years despite conventional reviews. As other AI firms race to develop Mythos‑class models, the urgency for coordinated standards, usage‑credit frameworks, and open‑source contributions grows. SKT’s involvement signals that leading telecoms are willing to invest heavily in AI‑enabled defenses, setting a benchmark for peers worldwide and shaping the next wave of cyber‑resilience strategies.