Vulnerability Exploitation Window Plunges to Two Hours as Taiwan Unveils AI Risk Framework

The proliferation of generative artificial intelligence and automated attack tools is driving a structural shift in the global cybersecurity threat landscape. Attackers are using AI to identify vulnerabilities in code and generate malware, posing a major challenge to traditional cyber defenses. Governments and businesses therefore urgently need robust AI governance and security management frameworks to protect critical information infrastructure and digital assets in an increasingly speed-driven contest.

According to the latest Zero Day Clock data, the average time from a vulnerability's disclosure to its exploitation plunged from 21.5 days in 2025 to two hours in July 2026. To guard against AI threats, Taiwan's Ministry of Digital Affairs announced its Artificial Intelligence Risk Classification Framework on July 7, 2026. Effective immediately, the framework divides AI risks into three main categories—technical flaws, operational interactions and societal impacts—spanning 20 subcategories.