# FBI FLASH: ARTEX intrusions reach privileged access in under 3 hours > The FBI's FLASH on ARTEX lists the tactics, indicators and mitigations for intrusions that run at machine speed. - **Severity:** Info: Research, tools, evals or lab claims with no direct exposure. Context, not an alert. - **Category:** Incident, Attack - **Published by:** Federal Bureau of Investigation - **Disclosed:** 2026-11-13 - **Affects:** ARTEX, ConnectWise Cloud Backup, Microsoft Entra ID - **Research by:** Federal Bureau of Investigation - **Primary source:** https://www.ic3.gov/CSA/2026/261113.pdf - **Page:** https://theinjection.dev/items/fbi-flash-artex-threat-overview/ The FBI published a TLP:CLEAR FLASH on the ARTEX agentic penetration testing framework, with indicators and mitigations. Some intrusions went from initial targeting to privileged access in under three hours. ## What happened The Federal Bureau of Investigation released FLASH-20261113-001 on 13 November 2026, a TLP:CLEAR alert on the ARTEX autonomous penetration testing framework. The FBI says malicious actors have used ARTEX since at least August 2026 against organizations worldwide across financial services, government, technology, defense, semiconductor, energy, healthcare, manufacturing, telecommunications and legal sectors. ## How it works ARTEX is a multi-agent platform written in Go that turns goals written in plain language into machine commands run by its own agents, with little human involvement after the objective is set. The framework ships no model of its own and calls external providers instead. The FBI's analysis of attacker-controlled instances found Chinese models most prevalent, led by the DeepSeek Flash family, which it attributes to their capability, low cost and lack of observable guardrails when a task is framed as penetration testing. ## Who is affected Organizations running internet-exposed applications, self-service portals, cloud infrastructure and partner or supplier systems are the FBI's stated focus. The alert singles out peripheral systems that receive less security attention than flagship applications: partner and supplier portals, mobile backends, employee applications, sales-support systems, legacy services and third-party integrations. ## What to do Inventory every internet-facing application, including the peripheral ones, and restrict cloud administration, publishing interfaces such as Kudu and SCM endpoints, hypervisors and backup platforms to trusted networks. Review Microsoft Entra audit records for new service principals, credential additions and consent grants, and hunt the indicators in the alert. Organizations that authorized the affected ConnectWise applications should review activity from at least 1 October 2026. ## Quick facts - **Alert:** FLASH-20261113-001, TLP:CLEAR - **Framework:** ARTEX, open-source, written in Go - **Activity since:** At least August 2026 - **Fastest observed:** Initial targeting to privileged access in under three hours - **Exposed consoles:** Over 1,600 as of early November 2026 - **Named campaigns:** Personal information on about 67,000 individuals at South Korean financial institutions; access to more than 2,250 ConnectWise Cloud Backup downstream tenants - **Models observed:** DeepSeek Flash family most prevalent; Claude, GPT, GLM, Grok and Kimi also supported ## Questions ### Is ARTEX still available now that the GitHub repository is gone? ARTEX remains available. The original developer, Autumn-27, removed the GitHub repository on 8 October 2026 after reports of threat actor abuse, but the FBI says the framework still spreads through community forks, localized variations, archived copies and a commercial version called ARTEX PRO. Researchers counted over 1,600 publicly exposed ARTEX consoles in early November 2026. ### Which models do attackers run ARTEX on? ARTEX includes no model of its own and integrates with external LLM providers. The FBI reports that Chinese models are most prevalent on attacker-controlled instances, with the DeepSeek Flash family dominating because of its balance of capability, relatively low cost and lack of observable guardrails when prompts are framed for penetration testing. Frontier models from Anthropic and OpenAI are also supported. ### What network signs point to ARTEX activity in my environment? Outbound calls to LLM endpoints including api.anthropic.com, api.deepseek.com, open.bigmodel.cn and api.x.ai are the clearest signal when they appear alongside other suspicious activity. The FBI also lists high-frequency automated requests with sequential enumeration against login and search endpoints, outbound transfers of structured personal data over HTTPS to unauthorized destinations, and .aspx or .jsp web shells in unexpected application paths. ### How does the FBI want ARTEX incidents reported? The FBI asks recipients to contact a local field office or the Internet Crime Complaint Center. For ARTEX cases specifically, the alert requests console screenshots and configuration files, service principal sign-in logs, Microsoft Graph and Entra audit records, outbound LLM endpoint connections, cloud and container activity, web shells and other persistence artifacts, and loss and impact figures. ## Sources 1. [FBI FLASH (PDF)](https://www.ic3.gov/CSA/2026/261113.pdf) 2. [GTIG: ARTEX Frameworks](https://cloud.google.com/blog/topics/threat-intelligence/artex-and-friends-enter-agentic-offensive-frameworks) 3. [Check Point: Autopilot](https://research.checkpoint.com/2026/hacking-on-autopilot-inside-artex-cybercrime-campaigns)