Kaspersky reports tenfold increase in AI service vulnerabilities

Kaspersky has reported a significant surge in vulnerabilities within artificial intelligence-based tools. According to data from the second quarter of 2026, 935 flaws were recorded in AI services, representing a tenfold increase compared to the fourth quarter of 2025.

Of these identified vulnerabilities, 119 were classified as critical, which is nearly five times the number recorded at the end of last year. The data, compiled via Kaspersky Vulnerability Management using databases such as CVE, the Russian BDU, and GitHub Advisory, reflects vulnerabilities that have been identified and made public rather than those confirmed to be exploited in active attacks.

The rapid increase is attributed to several factors: AI tools are being used more extensively to scan large volumes of code to find long-hidden issues, and the technology itself is increasingly being used to generate and correct software code, which sometimes bypasses adequate human verification.