"AI-Driven Cybersecurity Flaws Expected to Double This Year"

As artificial intelligence (AI) performance improves, the number of security vulnerabilities discovered in major software this year is expected to reach about twice the number identified last year.

According to Bloomberg News on the 27th (local time), the U.S. National Vulnerability Database (NVD), which compiles digital security flaws, has recorded 45,207 vulnerabilities from January through this date. This figure is already close to last year's total number of vulnerabilities registered for the entire year. Last year marked an all-time high in NVD-registered vulnerabilities.

Security vulnerabilities refer to software flaws that hackers can exploit to break into computer systems for criminal activities or espionage.

Major technology companies have also seen a dramatic increase in security vulnerabilities. Oracle fixed 1,449 vulnerabilities in its July regular software update. This is about 4.7 times more than the 309 vulnerabilities addressed in the same month last year. Microsoft also saw an increase to 642 vulnerabilities, about five times higher year-on-year. Google recently discovered and fixed 433 vulnerabilities in a Chrome update—about 39 times more than the 11 vulnerabilities fixed during the same update period last year.

Gabriel Bernadett-Shapiro, principal AI research scientist at cybersecurity firm SentinelOne, said, "We have to recognize the reality that these tools (AI) are strengthening people's ability to find software vulnerabilities."

Doug Turner, engineering lead for Google Chrome, said that vulnerabilities are being identified at an "unprecedented scale and speed" as a result of advances in AI models and greater investment that has followed.

However, the increase in vulnerability discoveries does not mean there has been a rise in actual cyberattacks such as hacking. According to the U.S. government, the number of vulnerabilities actively exploited this year has not increased. Of the 433 Chrome vulnerabilities found in July, 401 were identified internally by Google.

The ability of cutting-edge AI models to detect software vulnerabilities has advanced rapidly over the past several months. As a result, concerns are mounting that hackers could increasingly exploit such flaws.

In particular, advances in AI technology are accelerating the speed at which hackers carry out cyberattacks. Alexander Leslie, principal advisor at Recorded Future, stated that the average time it takes hackers to exploit a vulnerability in an actual attack has decreased from 72 hours last year to 24 hours this year.

This content was produced with the assistance of AI translation services.