AI speeds up vulnerability discovery and exploitation

Artificial intelligence is accelerating both vulnerability discovery and exploitation, shrinking the time security teams have to respond, according to a TechRadar article by Chris Wallis.

The article says Anthropic’s Mythos model and around 50 partner organizations reportedly identified more than 10,000 high- and critical-severity vulnerabilities in approximately one month. It also reports an estimated 6,202 high- and critical-severity vulnerabilities across more than 1,000 open-source projects.

Wallis cites 48,185 CVEs published in 2025, a 20.6% increase from the previous year. The article describes a researcher’s automated pipeline finding a remote, multi-stage SQL injection vulnerability affecting 300,000 WordPress plugins. It says the vulnerability was assigned CVE 2026 3985 and publicly disclosed.

The article also describes an operator called JadePuffer exploiting a vulnerable Langflow server, harvesting credentials, moving laterally, and encrypting 1,300 database records in an end-to-end ransomware operation.

Wallis argues that security teams need to reduce unnecessary internet exposure as well as shorten the time between finding and fixing vulnerabilities. He describes AI-assisted testing as a way to help teams validate findings and provide context, while noting that audit, insurance, and compliance frameworks will need to evolve. The article says continuous testing is emerging as standard practice.