Cybersecurity flaw discoveries soar to record high with use of AI: Data

In 2025, the number of publicly disclosed cybersecurity vulnerabilities surged to a record high, with over 48,000 cases logged, a two-and-a-half-fold increase from 2021. By June 2026, the count had risen to more than 35,000, underscoring the growing risks of cyberattacks and the urgent need for prompt fixes. This trend follows a string of Japanese companies reporting unauthorized access and data leaks, including the Times Car car-sharing service, leading the government to urge businesses to enhance security measures.

Cybersecurity vulnerabilities, once relatively scarce in the early 2010s, have seen a dramatic rise. Between 2017 and 2020, the organization recorded thousands of cases annually. However, with the advent of AI tools like OpenAI's ChatGPT, the pace of vulnerability discovery accelerated significantly. By 2024, the number of registered cases had reached 40,077 and hit 48,244 by 2025. This upward trend has continued into 2026, as AI models such as Anthropic’s Claude Mythos are employed to identify security flaws more rapidly.

Experts note that many companies are struggling to keep pace with these threats due to a shortage of skilled personnel in the cybersecurity field. Yukimi Sota, from the Japanese branch of US cybersecurity firm Proofpoint, highlighted this challenge. The discovery of cybersecurity vulnerabilities requires immediate action, as delays can leave systems exposed to unauthorized access and exploitation by malicious actors.

Mitre, a US nonprofit organization, compiles and shares these vulnerabilities to aid IT vendors and researchers in finding appropriate solutions.