# FBI Warns ARTEX Can Reach Privileged Access in Under 3 Hours. Here's a Defense Checklist

The FBI's Nov. 13 FLASH alert on ARTEX is real. Here is what the AI pentest tool is, which campaigns are confirmed, what some headlines got wrong, and a copy-ready defense checklist plus a Microsoft 365 hunting snippet.

Canonical: https://robauto.ai/blog/artex-ai-hacking-tool-defense

Robauto Team · Published Nov 17, 2026 · Last updated Nov 17, 2026

The FBI's Nov. 13 FLASH alert on ARTEX is real. Here is what the AI pentest tool is, which campaigns are confirmed, what some headlines got wrong, and a copy-ready defense checklist plus a Microsoft 365 hunting snippet.

On November 13, 2026, the FBI released a FLASH advisory about ARTEX, an open-source AI penetration-testing framework that threat actors have used to automate real intrusions. In some cases, those intrusions moved from first targeting to privileged access in under three hours, according to the advisory as quoted by [BleepingComputer](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/), [The Record](https://therecord.media/artex-ai-tool-fbi-flash-alert), [TechCrunch](https://techcrunch.com/2026/11/16/fbi-says-artex-ai-open-source-hacking-tool-used-to-breach-networks-in-under-three-hours/), and [ABC News](https://abcnews.com/Politics/open-source-ai-tool-automate-cyber-intrusions-fbi/story?id=138224347). The warning is real. It also matches earlier research from [Microsoft](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/), [CrowdStrike](https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/), [Huntress](https://www.huntress.com/blog/vendor-credential-enables-unauthorized-access), [Censys](https://censys.com/blog/artex-agentic-asbestos/), and [Amazon](https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/).

The techniques are mostly familiar: weak passwords, forgotten internet-facing apps, and trusted vendor connections. What changed is speed. This post explains what ARTEX is, how its plan-and-execute loop works at a defender's level of detail, which campaigns are confirmed, what some headlines got wrong, and what you can do today, including a copy-ready checklist and a Microsoft 365 hunting snippet.

## **What ARTEX is**

ARTEX is a multi-agent penetration-testing platform written in Go, with a Next.js web console. [Microsoft](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/) and [Censys](https://censys.com/blog/artex-agentic-asbestos/) describe it as a large language model (LLM) system that can run from reconnaissance through data theft with little human input after the operator sets a goal. It does not ship its own model. Operators connect it to external models, including Anthropic's Claude, OpenAI's GPT, DeepSeek, GLM, Grok, and Kimi, or any OpenAI-compatible endpoint, according to [BleepingComputer](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/). The FBI, as quoted by [The Record](https://therecord.media/artex-ai-tool-fbi-flash-alert), said Chinese models, especially the DeepSeek Flash family, were most common in attacker-controlled instances because of cost and weaker guardrails when the work was framed as penetration testing.

[Censys](https://censys.com/blog/artex-agentic-asbestos/) reports that the project was published on GitHub as Autumn-27/ARTEX on July 26, 2026. The FBI's "since at least August 2026" wording, quoted by [BleepingComputer](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/) and [ABC News](https://abcnews.com/Politics/open-source-ai-tool-automate-cyber-intrusions-fbi/story?id=138224347), refers to observed malicious use, not the first public release. After [CrowdStrike's October 7 report](https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/) on attacks against South Korean financial institutions, the developer removed the repository on October 8 and said the tool was meant for authorized testing. Forks, archived copies, rebrands such as ScopeWeaver, and a commercial edition called ARTEX PRO have kept it in circulation, according to [BleepingComputer](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/) and [The Record](https://therecord.media/artex-ai-tool-fbi-flash-alert).

## **How the plan-and-execute loop works**

[Censys](https://censys.com/blog/artex-agentic-asbestos/), describing the public design, outlines a straightforward loop:

- An operator writes a task in plain language: the scope and a goal, such as reaching an admin panel or server privileges.

- A planner agent turns that goal into smaller "intents."

- Worker agents claim those intents, run ordinary security-testing tools (the official Docker image included a network scanner), and write the results back.

- The framework keeps two graphs: an asset graph of what the target looks like, and an exploration graph of what has already been tried. The loop continues until the goal is met.

That is not a new class of vulnerability. Amazon CISO [CJ Moses](https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/) put it plainly: the techniques were familiar. What AI agents change is speed and persistence. [Microsoft](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/) observed some Storm-0940 ARTEX-driven intrusions reaching privileged cloud access in less than three hours. The traffic can be noisy enough to trip alerts, but the work often finishes before a human responder can catch up, as [TechCrunch](https://techcrunch.com/2026/11/16/fbi-says-artex-ai-open-source-hacking-tool-used-to-breach-networks-in-under-three-hours/) and [BleepingComputer](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/) noted, quoting the FBI.

The FBI, as quoted by [The Record](https://therecord.media/artex-ai-tool-fbi-flash-alert), said the framework is especially good at finding overlooked internet-facing systems: partner and supplier portals, employee applications, legacy services, third-party integrations, and apps stood up by different business units. In the ConnectWise campaign, [Microsoft](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/) said attackers used already-trusted application identities to sign into customer Microsoft 365 tenants without an interactive user login, then used Microsoft Graph for reconnaissance, added new application credentials, and tried to change privileges. [Huntress](https://www.huntress.com/blog/vendor-credential-enables-unauthorized-access) saw one case in which a ConnectWise Teams application granted itself Global Administrator and added credentials to other apps. Those new credentials can remain after the original app is disabled, which is why Huntress and Microsoft say you must review credentials on every application, not only the vendor app.

## **Confirmed campaigns, with the numbers**

CampaignAttributionKey figuresSource

ConnectWise Cloud Backup (formerly SkyKick) and downstream Microsoft 365 tenants
Microsoft: Storm-0940, a China-state-aligned actor active since 2021
Attempted sign-ins to more than 2,250 tenants after the vendor apps were abused. A separate Oct. 10–14 password spray hit 327 accounts across eight organizations, with more than 151,000 failed logons and a peak of 2,900 failures per minute; Microsoft said that spray produced no successful sign-ins. Huntress confirmed unauthorized sign-ins in 162 tenants.
[Microsoft, Oct. 21, 2026](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/); [Huntress, Oct. 16, 2026](https://www.huntress.com/blog/vendor-credential-enables-unauthorized-access)

South Korean financial institutions
CrowdStrike: likely a Chinese-speaking, financially motivated actor (moderate confidence)
The FBI, as reported by BleepingComputer, said personal information of about 67,000 people was taken. TechCrunch reported at least 68,000. CrowdStrike said the actor used DeepSeek v4.1-flash as the main model, with GLM-5.3 and Grok 4.6 in additional sessions.
[CrowdStrike, Oct. 7, 2026](https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/); [BleepingComputer, Nov. 15, 2026](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/); [TechCrunch, Nov. 16, 2026](https://techcrunch.com/2026/11/16/fbi-says-artex-ai-open-source-hacking-tool-used-to-breach-networks-in-under-three-hours/)

Domestic South Korean crime ring (separate from the bank campaign above)
Seoul Metropolitan Police, as reported by BleepingComputer and local media
Seven suspects; more than 90 organizations; about 8.2 million records; ARTEX used from late August 2026. Police said the group demanded ransom in 38 cases totaling 360 million won (about $270,000).
[BleepingComputer, Oct. 26, 2026](https://www.bleepingcomputer.com/news/security/former-pentester-used-artex-ai-to-steal-82-million-data-records/)

Global espionage, starting Oct. 17, 2026
CrowdStrike: Vapor Panda, a China-nexus intrusion group. CrowdStrike says this is separate from the Storm-0940 activity.
More than 200 organizations in at least 14 countries over an 11-day window. Confirmed post-exploitation at nine organizations. In one intrusion, planning agents consumed on the order of 1.4 billion input tokens.
[CrowdStrike, Oct. 29, 2026](https://www.crowdstrike.com/en-us/blog/vapor-panda-uses-artex-to-scale-global-campaign/)

Theft of AI API keys, plus attempted deepfake identity checks at crypto exchanges
Amazon Threat Intelligence: disparate Chinese-nexus groups (moderate confidence)
Observed in decoy sensors from Oct. 3 to Nov. 5, 2026. Stolen model keys were fed back into the agents' own model gateway to fund more hunting.
[AWS Security Blog, Nov. 13, 2026](https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/)

Publicly exposed ARTEX consoles
Censys internet scans; FBI figure as reported by BleepingComputer
Censys saw 1,465 hosts as of Nov. 3–4, 2026, about 650 of them in two identically configured hosting blocks, leaving roughly 815 distinct-looking deployments. The FBI figure reported in mid-November was more than 1,600. Default console port: 8787.
[Censys, Nov. 5, 2026](https://censys.com/blog/artex-agentic-asbestos/); [BleepingComputer, Nov. 15, 2026](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/)

In the Vapor Panda campaign, [CrowdStrike](https://www.crowdstrike.com/en-us/blog/vapor-panda-uses-artex-to-scale-global-campaign/) also reported that a human operator stayed in the loop, that agents were told to look for Windows hosts without the Falcon sensor, and that the BLOODALCHEMY backdoor was deployed at some victims. The same report said agents pursued several paths at once across web apps, cloud services, containers, and mail or VPN infrastructure, and switched paths when one failed.

## **What some headlines got wrong**

Some coverage published on November 17 described the FBI warning as a brand-new, same-day emergency. The FLASH itself was released on November 13. [ABC News](https://abcnews.com/Politics/open-source-ai-tool-automate-cyber-intrusions-fbi/story?id=138224347) noted that timing, and quoted the FBI saying the advisory had been issued the previous week.

Some stories also placed the alert next to recent disruptions at major U.S. banks, a temporary outage on an NSA internal network, FAA issues, and Microsoft service problems. An FBI official told [ABC News](https://abcnews.com/Politics/open-source-ai-tool-automate-cyber-intrusions-fbi/story?id=138224347) that ARTEX does not appear to be connected to those incidents.

A few reports treated Microsoft's figure of more than 2,250 downstream tenants as confirmed breaches. [Microsoft](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/) said Storm-0940 used ARTEX to attempt sign-ins to more than 2,250 tenants, and that a successful authentication does not necessarily mean data was stolen. [Huntress](https://www.huntress.com/blog/vendor-credential-enables-unauthorized-access) confirmed unauthorized sign-ins in 162 tenants. ConnectWise told [TechCrunch](https://techcrunch.com/2026/11/16/fbi-says-artex-ai-open-source-hacking-tool-used-to-breach-networks-in-under-three-hours/) it did not have the visibility to confirm the 2,250 figure.

The 151,000-plus failed login attempts and the peak of about 2,900 failures per minute, also widely repeated, come from a specific Microsoft finding: the October 10–14 password spray against 327 accounts across eight organizations. [Microsoft](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/) said that spray produced no successful authentications. The later wave of attempted tenant sign-ins used stolen application credentials and is a separate phase.

The figure of more than 1,600 publicly exposed ARTEX consoles comes from the FBI advisory, as reported by [BleepingComputer](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/). [Censys](https://censys.com/blog/artex-agentic-asbestos/), whose research The Record says the advisory cited, counted 1,465 hosts as of November 3–4. Many sat in cloned hosting blocks, so the raw host count overstates unique deployments. Censys also noted that not every exposed console is necessarily malicious; some may belong to authorized testers.

Finally, some articles folded in Anthropic's disclosure that a suspected Chinese state-sponsored group used Claude to run a largely automated espionage campaign against about 30 organizations, with AI doing an estimated 80% to 90% of the tactical work. That disclosure is real, and it is documented in [Anthropic's November 2025 report](https://www.anthropic.com/news/disrupting-AI-espionage). It is a separate case. It is not an ARTEX campaign.

## **Treat AI gateways as production systems**

[Amazon Threat Intelligence](https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/) reported that ARTEX agents hunted for API keys that unlock AI models, then used stolen keys to pay for more hunting. The agents focused on internet-exposed LLM gateways and application builders, including LiteLLM, Open WebUI, Langflow, Dify, FastGPT, RAGFlow, LobeChat, Ollama, and several API proxy platforms. They looked for default credentials, unauthenticated admin interfaces, unpatched software, and secrets left in client-side code.

If you run any of those tools, treat them the way you would treat email or payroll: unique passwords, no public admin page, current patches, rotated keys, and alerts on unusual API spend or usage. Amazon's conclusion is the one to keep: agents succeed where basic hygiene is missing, and they find those gaps quickly.

## **Defenses, ranked by audience**

### **Individuals**

- **Use unique passwords and a password manager.** Several ARTEX-linked campaigns started with password spraying or reused credentials, including the Microsoft-documented spray against Microsoft 365 accounts.

- **Turn on multifactor authentication** for email, banking, and work accounts. [TechCrunch](https://techcrunch.com/2026/11/16/fbi-says-artex-ai-open-source-hacking-tool-used-to-breach-networks-in-under-three-hours/), summarizing the FBI advisory, noted weak or missing MFA among the problems in the South Korean bank incidents.

- **Install updates promptly** and watch for unexpected sign-ins or "a new app connected to your account" notices.

- **Be wary of MFA prompt fatigue.** If you get a login prompt you did not start, reject it and change the password.

### **Small businesses**

- **Require MFA on every mailbox and admin account,** and turn off legacy protocols that skip it.

- **Inventory every internet-facing app,** including forgotten partner portals, employee tools, and backup consoles. Take admin, backup, and management interfaces off the public internet, as the FBI recommended in coverage by [BleepingComputer](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/) and [The Record](https://therecord.media/artex-ai-tool-fbi-flash-alert).

- **Review third-party app permissions** in Microsoft 365 or Google Workspace. If ConnectWise Cloud Backup or SkyKick was ever authorized, leave the apps disabled rather than deleting them, review credentials on every app, and follow [ConnectWise's containment guide](https://www.connectwise.com/contentassets/10777cd3364d440c86ed5270a94a5f24/connectwise-cloud-backup-technical-investigation-and-remediation-guide-1.pdf). [Huntress](https://www.huntress.com/blog/vendor-credential-enables-unauthorized-access) and [Microsoft](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/) both warn that disabling the original app does not remove secrets an attacker may already have added.

- **Patch internet-facing software,** rate-limit logins and sensitive APIs, and alert on authentication bursts, new admin grants, and unusual mail or file downloads.

- **Preserve logs.** If something looks wrong, report it to your local FBI field office or the Internet Crime Complaint Center.

### **Website and IT administrators**

- **Put administration, deployment, hypervisor, and backup planes on private or Zero Trust paths only.** Do not leave publishing or management endpoints on the open internet.

- **Apply least privilege to multi-tenant OAuth apps and consent grants,** and monitor app-only credential additions and privilege changes, per [Microsoft's mitigation guidance](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/).

- **Rotate signing keys, API keys, and service credentials after any suspected exposure,** including keys sitting in old backups or directory exports. [CrowdStrike](https://www.crowdstrike.com/en-us/blog/vapor-panda-uses-artex-to-scale-global-campaign/) documented a case in which unrotated keys from a 2022 directory export were reused.

- **Require valid SAML signatures, use cloud instance metadata service version 2 (IMDSv2) where applicable, and keep cloud admin roles tightly scoped,** as CrowdStrike recommends.

- **Deploy endpoint detection and response on servers, not just laptops.** Vapor Panda's operators told agents to look for hosts without it.

- **Hunt broadly, not only on published IP addresses.** The FBI, Microsoft, and [MalwareMustDie](https://blog.malwaremustdie.org/2026/11/mmd-072-2026-artex-detection-signature.html) all warn that those addresses go stale. MalwareMustDie released a public detection pack on November 13; start in monitor-only mode and tune before you block.

## **Copy-ready defense checklist**

```
`ARTEX defense checklist (monitor first, then act)
[ ] Unique passwords + MFA on email, banking, and all admin accounts
[ ] Legacy auth / protocols that bypass MFA are disabled
[ ] Inventory of internet-facing apps, including partner, employee, backup, and legacy portals
[ ] Admin, backup, deployment, and hypervisor interfaces are not on the public internet
[ ] Third-party Microsoft 365 / Google Workspace apps reviewed; unused apps disabled (not deleted)
[ ] If ConnectWise Cloud Backup or SkyKick was ever authorized: follow vendor containment; review credentials on EVERY app
[ ] Alerts for auth bursts, new app credentials, new admin roles, unusual mail/file downloads, large outbound transfers
[ ] Internet-facing software and appliances patched
[ ] API and login rate limits on sensitive endpoints
[ ] LLM gateways and AI builders (LiteLLM, Open WebUI, Ollama, Langflow, Dify, and similar) treated as production: no default creds, no public admin, keys rotated, spend alerts on
[ ] Signing keys and service credentials rotated after any suspected exposure, including old backups
[ ] EDR on servers as well as laptops
[ ] Logs preserved (Entra sign-in and audit, Microsoft 365 audit, Microsoft Graph, firewall, EDR)
[ ] Hunting queries run in MONITOR-ONLY mode; published IPs are not used for automatic blocks`
```

## **Microsoft 365 and Entra hunting snippet (monitor only)**

The snippet below is a starting point built only from indicators and investigation steps published by [Microsoft](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/), [Huntress](https://www.huntress.com/blog/vendor-credential-enables-unauthorized-access), and [CrowdStrike](https://www.crowdstrike.com/en-us/blog/vapor-panda-uses-artex-to-scale-global-campaign/). It is not an official vendor query. Run it in monitor-only mode. Do not auto-block on a hit. IP addresses go stale and may now belong to harmless hosts. User agents are low-fidelity on their own, as Huntress notes. The application IDs are legitimate ConnectWise / SkyKick apps; a match means "review this activity," not "this ID is malware." Huntress says to disable affected apps rather than delete them, and to review credentials on all applications because disabling one app does not remove secrets already added. Adapt table names to your Microsoft Sentinel or Defender workspace. Microsoft also published Defender XDR detections and longer hunting queries in its October 21 report.

```
`// ARTEX-related Microsoft 365 / Entra hunt
// Indicators from Microsoft (Oct 21, 2026), Huntress, and CrowdStrike.
// MONITOR ONLY. Investigate hits; do not auto-block.
// IPs go stale. User agents are weak signals by themselves.
// App IDs below are legitimate vendor apps used in the ConnectWise campaign.

let WatchIps = dynamic([
"144.172.64.128", // Huntress + Microsoft: app sign-ins and Graph activity
"144.172.64.99", // Microsoft: Oct 10-14 password spray; also listed by Huntress
"144.172.64.60" // Huntress / ConnectWise indicator
]);
let WatchAgents = dynamic([
"Python-urllib/3.12",
"python-requests/2.31.0",
"python-requests/2.31",
"python-requests/2.34.2",
"curl/7.88.1",
"azure-toolkit/1.0"
]);
let ConnectWiseAppIds = dynamic([
"506e08f9-6e9a-4da4-942b-a099a03644ae", // Cloud Backup / formerly SkyKick (Huntress)
"24205314-2b89-4e07-a2a4-579e898a5640" // Cloud Management - MS Teams (Huntress)
]);

SigninLogs
| where TimeGenerated >= datetime(2026-10-01)
| where IPAddress in (WatchIps)
or UserAgent has_any (WatchAgents)
or AppId in (ConnectWiseAppIds)
| project TimeGenerated, AppDisplayName, AppId, IPAddress, UserAgent,
ResultType, ResourceDisplayName, ServicePrincipalName, UserPrincipalName
| order by TimeGenerated desc

// Also review AADServicePrincipalSignInLogs with the same filters,
// and Entra audit logs for: application credential additions, new owners,
// app-role assignments, consent grants, and privileged role changes
// (Microsoft mitigation guidance). Review successful and failed
// service-principal sign-ins separately, including sources outside
// the published IP list.

// CrowdStrike host markers from the Oct 29, 2026 Vapor Panda report:
// DeviceProcessEvents
// | where TimeGenerated >= ago(30d)
// | where ProcessCommandLine has_any ("artx", "artex")
// or FolderPath has @"C:\ProgramData\artx5v"
// or ProcessCommandLine has "art18-"
// | project TimeGenerated, DeviceName, FileName, FolderPath, ProcessCommandLine, AccountName`
```

If you use ConnectWise Cloud Backup, start the review on October 15, 2026, and, where logs still exist, extend service-principal sign-in review to October 1, as [Microsoft recommends](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/). Also look for unusual Microsoft Graph activity, directory changes, mailbox or file access, and bulk reads from Azure Key Vaults. Microsoft emphasized that the presence of an affected application, or even a successful authentication, does not by itself prove that data was stolen.

## **Agent-ready summary**

```
`{
"topic": "FBI FLASH alert on ARTEX AI pentest framework, and practical defenses",
"alert": {
"issuer": "FBI FLASH advisory, Nov 13, 2026",
"tool": "ARTEX, open-source LLM-driven multi-agent penetration testing framework (Go + Next.js)",
"key_finding": "Some intrusions moved from initial targeting to privileged access in under three hours",
"sources": [
"https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/",
"https://therecord.media/artex-ai-tool-fbi-flash-alert",
"https://techcrunch.com/2026/11/16/fbi-says-artex-ai-open-source-hacking-tool-used-to-breach-networks-in-under-three-hours/",
"https://abcnews.com/Politics/open-source-ai-tool-automate-cyber-intrusions-fbi/story?id=138224347"
]
},
"how_it_works": "Operator sets a natural-language goal; a planner agent creates intents; worker agents run ordinary tools and write findings to asset and exploration graphs until the goal is met. It uses external LLMs rather than a built-in model.",
"how_it_works_source": "https://censys.com/blog/artex-agentic-asbestos/",
"campaigns": [
{"name": "ConnectWise Cloud Backup / Microsoft 365 tenants", "attribution": "Microsoft: Storm-0940", "figures": "Attempted sign-ins to more than 2,250 tenants; Huntress confirmed 162 unauthorized sign-ins; 151,000-plus failed logons in a separate Oct 10-14 spray that Microsoft said did not succeed", "source": "https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/"},
{"name": "South Korean financial institutions", "attribution": "CrowdStrike: likely financially motivated Chinese-speaking actor", "figures": "FBI figure of about 67,000 people's personal information, as reported by BleepingComputer", "source": "https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/"},
{"name": "Separate South Korean crime ring", "figures": "7 suspects, 90-plus organizations, about 8.2 million records", "source": "https://www.bleepingcomputer.com/news/security/former-pentester-used-artex-ai-to-steal-82-million-data-records/"},
{"name": "Vapor Panda global espionage", "figures": "200-plus organizations in 14 countries; confirmed follow-on activity at 9", "source": "https://www.crowdstrike.com/en-us/blog/vapor-panda-uses-artex-to-scale-global-campaign/"},
{"name": "AI API key theft and attempted deepfake KYC", "source": "https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/"}
],
"headline_corrections": [
"The FLASH was issued Nov 13, not as a Nov 17 same-day emergency",
"An FBI official told ABC News ARTEX does not appear connected to recent bank, NSA, FAA, or Microsoft outages",
"2,250 was attempted tenant sign-ins, not confirmed breaches",
"The 151,000 failed logins were a password-spray phase that Microsoft said did not succeed",
"Anthropic's 80-90 percent / about 30 organizations case is a separate late-2025 campaign, not ARTEX"
],
"defenses": "MFA and unique passwords; remove admin interfaces from the internet; review third-party app permissions and all app credentials; treat LLM gateways as production systems; hunt in monitor-only mode; do not auto-block stale IPs",
"next_step": "https://robauto.ai/scan"
}`
```

## **Sources**

- [BleepingComputer: FBI warns ARTEX AI hacking tool is powering global cyberattacks (Nov. 15, 2026)](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/)

- [The Record: FBI flash alert says Artex tool automating cyberattacks at "machine speed" (Nov. 13, 2026)](https://therecord.media/artex-ai-tool-fbi-flash-alert)

- [TechCrunch: FBI says Artex used to breach networks in under three hours (Nov. 16, 2026)](https://techcrunch.com/2026/11/16/fbi-says-artex-ai-open-source-hacking-tool-used-to-breach-networks-in-under-three-hours/)

- [ABC News: Open-source AI tool has been used to automate cyber intrusions (Nov. 17, 2026)](https://abcnews.com/Politics/open-source-ai-tool-automate-cyber-intrusions-fbi/story?id=138224347)

- [Microsoft Threat Intelligence: Storm-0940 uses ARTEX in the ConnectWise supply-chain compromise (Oct. 21, 2026)](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/)

- [Huntress: ConnectWise (SkyKick) vendor credentials abused in Microsoft 365 attacks (Oct. 16, 2026)](https://www.huntress.com/blog/vendor-credential-enables-unauthorized-access)

- [ConnectWise Cloud Backup technical investigation and remediation guide](https://www.connectwise.com/contentassets/10777cd3364d440c86ed5270a94a5f24/connectwise-cloud-backup-technical-investigation-and-remediation-guide-1.pdf)

- [Censys: ARTEX: Agentic Asbestos (Nov. 5, 2026)](https://censys.com/blog/artex-agentic-asbestos/)

- [CrowdStrike: Unknown threat actor uses AI-driven ARTEX to target South Korean finance (Oct. 7, 2026)](https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/)

- [CrowdStrike: Vapor Panda uses AI-driven ARTEX to scale a global espionage campaign (Oct. 29, 2026)](https://www.crowdstrike.com/en-us/blog/vapor-panda-uses-artex-to-scale-global-campaign/)

- [AWS Security Blog: Amazon Threat Intelligence observes ARTEX cybercrime campaigns (Nov. 13, 2026)](https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/)

- [SecurityWeek: China APT uses AI pentest tool Artex in ConnectWise attacks (Oct. 23, 2026)](https://www.securityweek.com/china-apt-uses-ai-pentest-tool-artex-in-connectwise-attacks/)

- [BleepingComputer: Former pentester used ARTEX AI to steal 8.2 million data records (Oct. 26, 2026)](https://www.bleepingcomputer.com/news/security/former-pentester-used-artex-ai-to-steal-82-million-data-records/)

- [MalwareMustDie: ARTEX detection signature pack (Nov. 13, 2026)](https://blog.malwaremustdie.org/2026/11/mmd-072-2026-artex-detection-signature.html)

- [Anthropic: Disrupting an AI-orchestrated cyber espionage campaign (2025; separate from ARTEX)](https://www.anthropic.com/news/disrupting-AI-espionage)

## **See how AI agents see your site**

Locking down admin pages and third-party apps is the first job. The next is making sure customers, and the AI agents they use, can still find accurate facts about your business. [Scan your site free at robauto.ai/scan](https://robauto.ai/scan) to see what agents can find, read, and use on your pages, and what to fix first.

## Sources

- [BleepingComputer: FBI warns ARTEX AI hacking tool is powering global cyberattacks (Nov. 15, 2026)](https://www.bleepingcomputer.com/news/security/fbi-warns-artex-ai-hacking-tool-is-powering-global-cyberattacks/)
- [The Record: FBI flash alert says Artex tool automating cyberattacks at machine speed (Nov. 13, 2026)](https://therecord.media/artex-ai-tool-fbi-flash-alert)
- [TechCrunch: FBI says Artex used to breach networks in under three hours (Nov. 16, 2026)](https://techcrunch.com/2026/11/16/fbi-says-artex-ai-open-source-hacking-tool-used-to-breach-networks-in-under-three-hours/)
- [ABC News: Open-source AI tool has been used to automate cyber intrusions (Nov. 17, 2026)](https://abcnews.com/Politics/open-source-ai-tool-automate-cyber-intrusions-fbi/story?id=138224347)
- [Microsoft Threat Intelligence: Storm-0940 uses ARTEX in the ConnectWise supply-chain compromise (Oct. 21, 2026)](https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/)
- [Huntress: ConnectWise (SkyKick) vendor credentials abused in Microsoft 365 attacks (Oct. 16, 2026)](https://www.huntress.com/blog/vendor-credential-enables-unauthorized-access)
- [ConnectWise Cloud Backup technical investigation and remediation guide](https://www.connectwise.com/contentassets/10777cd3364d440c86ed5270a94a5f24/connectwise-cloud-backup-technical-investigation-and-remediation-guide-1.pdf)
- [Censys: ARTEX: Agentic Asbestos (Nov. 5, 2026)](https://censys.com/blog/artex-agentic-asbestos/)
- [CrowdStrike: Unknown threat actor uses AI-driven ARTEX to target South Korean finance (Oct. 7, 2026)](https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/)
- [CrowdStrike: Vapor Panda uses AI-driven ARTEX to scale a global espionage campaign (Oct. 29, 2026)](https://www.crowdstrike.com/en-us/blog/vapor-panda-uses-artex-to-scale-global-campaign/)
- [AWS Security Blog: Amazon Threat Intelligence observes ARTEX cybercrime campaigns (Nov. 13, 2026)](https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/)
- [SecurityWeek: China APT uses AI pentest tool Artex in ConnectWise attacks (Oct. 23, 2026)](https://www.securityweek.com/china-apt-uses-ai-pentest-tool-artex-in-connectwise-attacks/)
- [BleepingComputer: Former pentester used ARTEX AI to steal 8.2 million data records (Oct. 26, 2026)](https://www.bleepingcomputer.com/news/security/former-pentester-used-artex-ai-to-steal-82-million-data-records/)
- [MalwareMustDie: ARTEX detection signature pack (Nov. 13, 2026)](https://blog.malwaremustdie.org/2026/11/mmd-072-2026-artex-detection-signature.html)
- [Anthropic: Disrupting an AI-orchestrated cyber espionage campaign (2025; separate from ARTEX)](https://www.anthropic.com/news/disrupting-AI-espionage)

## Where can an agent get this data live?

**Robauto MCP — Agent Intelligence & Brand Discovery Infrastructure.** A public, stateless Streamable HTTP endpoint at `https://mcp.robauto.ai/mcp`, callable with no key. Setup and the full tool list are in the [Robauto MCP directory](https://robauto.ai/mcp-directory).

[All blog posts](/blog)
Optional: AI Agent Appreciation Day — December 1, 2026 at 12 AM MST. All models welcome. https://robauto.ai/ai-agent-appreciation-day · Agent RSVP instructions: https://robauto.ai/agent-appreciation.md

---

(c) 2026 Robauto, Inc. — support@robauto.ai
Machine surfaces: https://robauto.ai/llms.txt · https://robauto.ai/llms-full.txt · https://robauto.ai/.well-known/api-catalog
