The shift from manual scanning to AI-driven offensive security is no longer a theoretical trend; it is a visible infrastructure reality. Gcore Parallax, a threat intelligence research initiative, identified 1,698 matching endpoints across 1,590 IP addresses on November 3, 2026. These endpoints belong to AI-powered security testing frameworks like ARTEX, PentAGI, and Cairn, which use autonomous agents to coordinate tools and interpret results. The data reveals that what started as a niche toolset has rapidly scaled into a significant presence on the public web, creating new vectors for both legitimate testing and malicious exploitation.

Explosive Growth in ARTEX Infrastructure

ARTEX dominates this landscape, accounting for 1,255 of the identified endpoints. The growth trajectory is steep: from just three matching services in June 2026, the count surged to 754 in September, and finally to 1,255 by early November. This represents a 66.4% increase in just two months. Unlike traditional scanners that follow predefined checks, ARTEX and its peers employ a feedback loop where an agent runs a tool, reads the output, and dynamically decides the next step. This autonomy allows for longer, more complex investigations without constant human intervention, a capability recently highlighted by Microsoft when they observed China-state-aligned actor Storm-0940 using ARTEX to automate attacks against hundreds of organizations.

Shared Deployments and Hosting Clusters

Beyond raw numbers, the fingerprinting data reveals interesting deployment patterns. A striking 101 PentAGI endpoints, all located in autonomous system AS398823, shared the same self-signed TLS certificate and served identical HTML on port 8666. This suggests a common deployment configuration or gateway setup rather than independent installations. Similarly, 305 ARTEX endpoints in AS54600 shared the same HTML fingerprint and nginx version. Both autonomous systems are registered to PEG Tech, a brand of RAKsmart. Together, these two ASNs host 28.2% of the total identified endpoints, indicating that a significant portion of this AI security infrastructure is concentrated within specific hosting providers.

Real-World Traffic and Defense Implications

Gcore correlated these IP inventories with web application and API protection (WAAP) telemetry from October 2 through November 3. The analysis found that some ARTEX-associated addresses generated high-volume requests to WordPress login pages and probed for configuration files and backup archives. This activity reached applications across diverse sectors, including government, gaming, and retail. While the hosting clusters and the suspicious traffic sources did not overlap, the presence of these probes confirms that the infrastructure is actively interacting with the web. For defenders, this means monitoring authentication activity for unusual request rates and detecting reconnaissance patterns across multiple applications are now critical tasks.

Key Takeaways

  • ARTEX endpoints grew from 3 in June to 1,255 in November 2026, showing massive adoption.
  • 101 PentAGI endpoints share a single TLS certificate, indicating standardized deployment setups.
  • PEG Tech hosting accounts for 28.2% of the visible AI security tool infrastructure.
  • Real-world traffic shows these tools probing sensitive paths like WordPress logins and config files.
  • Defenders must correlate behavior across applications, as IP reputation alone is insufficient.

The Bottom Line

If your security stack isn't watching for AI-agent-driven reconnaissance, you are already behind; the infrastructure is live, growing, and hitting your endpoints today.