Mythos Marks New Era in Global AI Security

Anthropic's 'Mythos', boasting top-tier AI security detection capabilities, is emerging as a 'double-edged sword' for Wall Street banks amid U.S. Treasury recommendations, raising concerns about global AI security. Anthropic's new AI model 'Mythos' recorded 83.1% in a CyberGym evaluation, significantly outperforming the previous top model, Opus 4.6 (66.6%) in security detection. It is known for finding a 27-year-old bug in OpenBSD and a 16-year-old vulnerability that automated tools failed to detect after over 5 million checks. While the U.S. Treasury has recommended the use of the 'Mythos' model to major Wall Street banks such as Goldman Sachs, Citi, Bank of America, and Morgan Stanley, its potential for hackers to exploit it for fully automated infiltration has led to its evaluation as a 'double-edged sword'.

These AI security issues are also being prioritized by the Trump administration, drawing attention as figures like Vice President JD Vance hold closed-door meetings with CEOs of major AI companies.

Meanwhile, the adoption of digital twin technology for site management and process optimization is accelerating, emerging as essential infrastructure for manufacturing innovation. Kia's Gwangmyeong EVO plant significantly reduced the effort required to identify equipment anomalies from 120 minutes to 10 minutes by implementing an integrated digital twin control system, while Samsung Electronics and SK Hynix are leveraging NVIDIA's 'Omniverse' platform to optimize semiconductor production.

In parallel, the global energy and industrial structure reform competition is intensifying. China is operationalizing industrial plans spanning decades to enhance its energy self-sufficiency, and Japan's major companies are collaborating to develop domestic AI. South Korea, on the other hand, is navigating a period of simultaneous preparation for a surge in electricity demand by 2040 and restructuring its petrochemical industry.