Cloudflare tracks 50 AI pentest frameworks scanning customers
Cloudflare says more than 50 families of AI penetration testing frameworks tried to scan its customers in three days. Daily requests from these tools are up 30 times since October 1, and the company lists six defences.
By JetherVerse News

Cloudflare reports a sharp rise in AI penetration testing frameworks aimed at its customers. In the last three days it saw more than 50 framework families try to scan customers around the world, and the number of IP addresses running them is in the thousands and still growing.
Cloudflare traces much of this to ARTEX, an open source agentic penetration testing system. Its creator, who goes by Autumn-27, took it down on October 8, saying the malicious use of the tool ran against its original intent. The code lives on. There are now more than 12,000 repositories, with several hundred forks carrying lineage from the original. Cloudflare says daily requests from AI penetration testing and exploitation tools are up 30 times since October 1. These frameworks plug into Model Context Protocol servers and standalone security tools, map an environment as a graph and run parallel agents to enumerate, test and attack at the speed of inference. Some chain stolen customer credentials to move sideways inside a network. Cloudflare is tracking more than 50 such agents, and the top five each have over 1,000 GitHub stars and run fully on their own.
The change is simple. Skill no longer limits the attacker. An automated framework observes, picks the most vulnerable target, chooses an action and acts, then repeats. To meet this, Cloudflare recommends what it calls the Modern Security Model, with six pillars. Put a web application firewall with rate limits at the front. Run anomaly detection on every web application your customers reach. Use bot defense to spot phony visitors. Close the back door by gating high-risk internal and external apps behind identity. Enforce validated schemas and explicit security controls on what you accept. Gather threat intelligence so you see attacks early.
Small businesses in Nigeria with public web apps and weak login screens are as exposed as large ones, since an automated scanner does not pick targets by size. A rate limit and a login check cost little compared with a breach.
If you build or host web apps for clients, review these six pillars against your own setup this week.
Source: Cloudflare Blog
Get the news in your inbox
The tech stories that matter from JetherVerse News, with the source on every one.

