1,465 hosts worldwide are exposing an ARTEX console to the public internet, Censys security researcher Silas Cutler reported on November 5 — a scale that, the data suggest, puts ARTEX far beyond the handful of incidents first flagged by security firms.
ARTEX's design, release and rapid redistribution
Released on GitHub in July 2026 by a user identified as “Autumn-27,” ARTEX is an open-source, AI model–driven autonomous penetration-testing framework written in Go with an embedded Next.js web console. The framework presents itself as an offensive command-and-control (C2) platform that communicates with several AI agents and can be plugged into large language models (LLMs).
The original repository was removed by its creator in October 2026 after widespread abuse, the author said; the creator also asserted they were not involved in the ARTEX-linked attacks and that the tool was intended for authorized penetration testing. Despite the takedown, ARTEX has continued to spread through community forks and derivatives, including localized variants such as ScopeWeaver and a commercial product called ARTEX PRO.
Three tracked campaigns tied to ARTEX
Security firms and national authorities have linked ARTEX to multiple malicious operations since October 2026. In early October, CrowdStrike reported that ARTEX had been used alongside Anthropic’s Claude AI model in a campaign targeting South Korean financial institutions; CrowdStrike assessed with moderate confidence the attacker was a Chinese speaker and financially motivated, and said the activity ran from late September to early October.
Later in October, Microsoft traced ARTEX to a large-scale compromise affecting ConnectWise customer tenants that forced the IT management software provider to take several cloud services offline. Microsoft attributed that campaign to a China-state-aligned actor tracked as Storm-0940 (also known as CovertNetwork-1658 or ORB07).
On November 4, India’s Computer Emergency Response Team (CERT‑In) warned that a group tracked as “Crafty Falcon” had been targeting financial and cryptocurrency organizations in successive attacks using several AI-powered tools, including ARTEX.
This site is the portfolio.
OSINTSights runs on Cloudflare Workers, D1, R2, and Vectorize, with an AI pipeline on Hetzner ARM. Nubivance designed, built, and operates it. We do the same for clients.
See what we buildCensys scan: footprint, clustering, and co-hosted threats
The Censys report provides the first public quantitative view of ARTEX’s global distribution. Cutler found 1,465 internet-exposed ARTEX consoles; roughly 650 of those sit “within two blocks of identically configured hosting,” while about 815 deployments look distinct. That distribution implies hundreds of separate deployments rather than a small number of cloned instances.
Censys telemetry showed 155 ARTEX hosts — about 10.6% — also ran at least one other artifact Censys categorizes as a threat. Examples cited include Vshell, Cobalt Strike and Adaptix C2, as well as other AI attack frameworks, reconnaissance platforms and tunneling tools. The asset-discovery and vulnerability-scanning platform ScopeSentry, maintained by ARTEX’s creator, is also listed by Censys as a threat artifact.
Shared AI tooling, accessible toolkits, and model gateways
Censys found ARTEX frequently sharing infrastructure with other AI-powered penetration-testing frameworks such as CyberStrikeAI, PentAGI, HexStrike AI and RedAmon; CyberStrikeAI appeared alongside ARTEX on 70 hosts. Cutler noted, “Operators appear to be using multiple agent frameworks side by side.”
The researchers identified 32 ARTEX hosts with publicly accessible directories containing tools used across stages of cyber-attacks — post-exploitation kits, stagers and implants — and found 43 ARTEX hosts running self-hosted LLM gateway or reseller software. Those findings underscore that access to AI models has become another component of the offensive infrastructure stack, not simply an add‑on.
Independent research from Check Point, cited in Censys’ reporting, showed AI agents in ARTEX control panels can run large-scale attacks with relatively little human input for “just a few hundred dollars,” though these automated chains still fail or “hallucinate” successful outcomes.
What this means for financial and cryptocurrency organizations, security teams, and CERTs
- Financial and cryptocurrency organizations: CERT‑In’s November 4 advisory directly named these sectors as targets; the Censys findings reinforce that ARTEX-powered infrastructure is publicly reachable at scale and has been observed alongside tools used in post‑exploitation, increasing the risk profile for these organizations.
- Security operations and incident responders: The clustering of ARTEX with other C2 frameworks and AI agent toolsets suggests defenders must look for mixed-tool campaigns and public-facing consoles or directories that expose stages of compromise. The presence of self-hosted LLM gateways indicates model access can be part of the adversary’s supply chain to track and disrupt.
- National CERTs and platform owners: The GitHub takedown of the original ARTEX repository did not halt proliferation; community forks and commercialized variants continue to appear. CERTs and platform operators will face repeated containment decisions about whether takedowns, blocking, or notification to hosting providers will be effective against a distributed, forked footprint.
The record compiled by Censys, alongside the earlier attributions from CrowdStrike, Microsoft and CERT‑In, sketches a rapid arc: an open-source offensive framework released for penetration testing, weaponized in multiple campaigns, forked and commercialized, and now visible on more than a thousand public hosts — some sharing hosting with recognized C2 tooling and self-hosted model gateways. The key, immediate question the facts leave: with hundreds of distinct-looking deployments and LLM access baked into offensive stacks, can takedowns of single repositories or individual hosts materially slow the spread?
