Cloudflare says more than 50 families of AI attack frameworks tried to scan its customers, with ARTEX and its forks over 60%

Cloudflare · Company report

Read the source (opens in a new tab)
What happened
Cloudflare published an analysis on November 13, 2026 of AI penetration testing orchestrators seen against its customers, with findings from its Cloudforce One incident response team.
Key points
  • Over three days it saw more than 50 framework families try to scan Cloudflare customers; traffic from such tools has grown more than 30x since October 1.
  • After ARTEX's creator removed it on October 8, Cloudflare counts over 12,000 repositories and several hundred forks, and calls the jump in instances a likely "hydra effect."
  • In one case ARTEX used a stolen CI/CD API token with no IP restrictions to create many Worker scripts and send 400 database dump requests in an hour.
  • Cloudflare says attackers repeatedly took static, non-expiring CI/CD tokens, and believes much of that case's activity was "lightly guided by the human operator."
Evidence
Company report from a network provider based on its traffic telemetry and its own incident response cases. Customers are not named, and the defenses it recommends include its own products.
Uncertain
Attribution of traffic to specific frameworks is Cloudflare's own; it says shares overlap because one IP address can run several tools.

See it in the latest record