Experts face growing challenges: AI uncovers thousands of security vulnerabilities

Artificial intelligence technologies have helped discover thousands of security vulnerabilities in recent months, placing cybersecurity experts before increasing challenges. Since last May, OpenAI models have helped detect over 1,300 vulnerabilities, including two in the PHP language that remained hidden for decades.

Artificial intelligence technologies have helped discover thousands of security vulnerabilities in recent months, placing cybersecurity experts before increasing challenges to keep up with repair operations. According to a report by The New York Times, since last May, when OpenAI began using its models to scan open-source software, these technologies have helped detect over 1,300 security vulnerabilities, prompting the company to allocate teams and significant computing capabilities to follow up on discovered issues and assist in addressing them. Among the most notable discoveries contributed by the company's models are two vulnerabilities in the PHP programming language, which remained hidden for decades and could have affected numerous websites worldwide, including government sites. A cybersecurity preparedness and safety official at OpenAI, Fuad Matin, described in his conversation with the newspaper what is happening as akin to opening Pandora's box of security vulnerabilities, referring to the scale of problems that AI tools have become able to uncover. Matin said that on one occasion he had to return to work after only two hours of sleep on his birthday night, due to the pressures resulting from following up on discovered vulnerabilities.