# Unit 42: 1,323 hosts run CyberStrikeAI and 609 run ARTEX

> Unit 42 counts the agentic pentest frameworks exposed on the internet and gives defenders timing patterns to hunt for.

- **Severity:** Info: Research, tools, evals or lab claims with no direct exposure. Context, not an alert.
- **Category:** Research, Incident
- **Published by:** Unit 42
- **Disclosed:** 2026-11-11
- **Affects:** ARTEX, CyberStrikeAI, Cairn, PentAGI
- **Research by:** Cuong Dinh, Richard Emerson, Samantha Stallings
- **Primary source:** https://unit42.paloaltonetworks.com/artex-and-ai-powered-red-teaming-tools/
- **Page:** https://theinjection.dev/items/unit42-artex-red-team-tool-census/

Unit 42 counted 1,323 internet-facing hosts running CyberStrikeAI and 609 running ARTEX as of October 2026, and published the timing patterns its responders use to tell agent-driven intrusions from human ones.

## What happened

Unit 42 published a defender guide to ARTEX and AI-powered red teaming tools on November 11, 2026, written by Cuong Dinh, Richard Emerson and Samantha Stallings. The team counted internet-facing hosts running the two most common frameworks and analyzed 14 tools across the ecosystem. The guide draws on Unit 42 incident response and managed threat hunting cases, including intrusions where the attacker's server hosted an ARTEX console.

## How it works

Agent-driven intrusions leave a timing signature, because the operator's tooling waits while the model thinks. Unit 42 observed bursts of one to 14 requests about a second apart, followed by pauses of 25 to 31 seconds. In one case more than 200 client secrets were validated in about three minutes, and one host polled every 48 seconds for three days. Operators also used bogus canary usernames and placeholder values before real attempts.

## Who is affected

Organizations with public attack surface are the targets, in particular those exposing OpenAPI specifications, source maps, and configuration files holding tokens or client IDs, plus self-service registration and password reset flows. Unit 42 reports that some intrusions reached cloud access through public password reset pages and exposed APIs without exploiting any software vulnerability, sometimes in under a day.

## What to do

Remove public OpenAPI specs, source maps and config files carrying tokens or client IDs, and require authentication on every write endpoint, including dev and staging. Rate-limit and bot-protect registration, password reset and verification flows, and drop unauthenticated account-existence checks. Hunt for behavior rather than indicators: canary usernames, a cloud key's first use from a new IP, and write-then-delete storage tests.

## Quick facts

- **CyberStrikeAI hosts:** 1,323 internet-facing, as of October 2026
- **ARTEX hosts:** 609 internet-facing, 21 hosts ran both frameworks
- **Tools analyzed:** 14, with 12 emerging tools highlighted
- **Tool categories:** Recon and attack surface, agent-led testing, source-code auditing, AI-system assessment
- **Timing signature:** Bursts of 1 to 14 requests, then pauses of 25 to 31 seconds
- **Found by:** Cuong Dinh, Richard Emerson, Samantha Stallings
- **Published:** 2026-11-11

## Sources

1. [Unit 42 Research](https://unit42.paloaltonetworks.com/artex-and-ai-powered-red-teaming-tools/)
2. [Unit 42 Cairn Report](https://unit42.paloaltonetworks.com/cairn-framework-ai-autonomous-attacks/)